Lactate-mediated epigenetic reprogramming regulates formation of human pancreatic cancer-associated fibroblasts

  1. Tushar D Bhagat
  2. Dagny Von Ahrens
  3. Meelad Dawlaty
  4. Yiyu Zou
  5. Joelle Baddour
  6. Abhinav Achreja
  7. Hongyun Zhao
  8. Lifeng Yang
  9. Brijesh Patel
  10. Changsoo Kwak
  11. Gaurav S Choudhary
  12. Shanisha Gordon-Mitchell
  13. Srinivas Aluri
  14. Sanchari Bhattacharyya
  15. Srabani Sahu
  16. Prafulla Bhagat
  17. Yiting Yu
  18. Matthias Bartenstein
  19. Orsi Giricz
  20. Masako Suzuki
  21. Davendra Sohal
  22. Sonal Gupta
  23. Paola A Guerrero
  24. Surinder Batra
  25. Michael Goggins
  26. Ulrich Steidl
  27. John Greally
  28. Beamon Agarwal
  29. Kith Pradhan
  30. Debabrata Banerjee
  31. Deepak Nagrath  Is a corresponding author
  32. Anirban Maitra  Is a corresponding author
  33. Amit Verma  Is a corresponding author
  1. Albert Einstein College of Medicine, Montefiore Medical Center, United States
  2. University of Michigan, United States
  3. Rutgers University, United States
  4. UT MD Anderson Cancer Center, United States
  5. Cleveland Clinic, United States
  6. University of Nebraska Medical Center, United States
  7. Johns Hopkins, United States
  8. GenomeRxUs LLC, United States

Abstract

Even though pancreatic ductal adenocarcinoma (PDAC) is associated with fibrotic stroma, the molecular pathways regulating the formation of cancer associated fibroblasts (CAFs) are not well elucidated. An epigenomic analysis of patient-derived and de-novo generated CAFs demonstrated widespread loss of cytosine methylation that was associated with overexpression of various inflammatory transcripts including CXCR4. Co-culture of neoplastic cells with CAFs led to increased invasiveness that was abrogated by inhibition of CXCR4. Metabolite tracing revealed that lactate produced by neoplastic cells leads to increased production of alpha-ketoglutarate (aKG) within mesenchymal stem cells (MSCs). In turn, aKG mediated activation of the demethylase TET enzyme led to decreased cytosine methylation and increased hydroxymethylation during de novo differentiation of MSCs to CAF. Co-injection of neoplastic cells with TET-deficient MSCs inhibited tumor growth in vivo. Thus, in PDAC, a tumor-mediated lactate flux is associated with widespread epigenomic reprogramming that is seen during CAF formation.

https://doi.org/10.7554/eLife.50663.001

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is a deadly disease and is the third leading cause of deaths from cancer in the United States. An exuberant host fibrotic response, termed stromal desmoplasia, is a characteristic feature of PDAC (Feig et al., 2013; Biffi et al., 2019), (Feig et al., 2012; Yu et al., 2012). The stromal fibrosis is suspected to contribute to chemoresistance in pancreatic cancer by impeding drug delivery (Provenzano et al., 2012; Öhlund et al., 2017; von Ahrens et al., 2017), and differences in stromal behavior have been implicated with patient outcomes in pancreatic and other cancers. The stromal microenvironment predominantly consists of cancer-associated fibroblasts (CAFs) that are activated during tumorigenesis, undergoing morphological and functional changes, when compared to normal fibroblasts (Kalluri, 2016). CAFs are derived via activation of resident pancreatic stellate cells, and also from differentiation and activation of bone marrow derived mesenchymal stem cells (BM-MSCs) that migrate to the peritumoral milieu due to chemotactic signals released by cancer cells. The characteristic features of activated CAFs include the expression of α-smooth-muscle actin (α-SMA (ACTA1)) (Öhlund et al., 2017), enhanced secretory and contractile ability, increased synthesis of extracellular matrix proteins, such as collagens, and of growth factors including basic fibroblast growth factor, transforming growth factor beta (TGF-β), interleukin-8 and platelet-derived growth factors (PDGF). In preclinical models of PDAC (Feig et al., 2012; Hwang et al., 2008; Xu et al., 2010), as well as in other cancer types (Karnoub et al., 2007), CAFs have been shown to promote invasion and metastases through myriad mechanisms. Multiple lines of evidence support the existence of robust paracrine signals from neoplastic epithelial cells to the stromal compartment (Behrens et al., 2010; Bailey et al., 2009; Omary et al., 2007), which likely facilitates the reprogramming of MSCs to an activated CAF-like state, that in turn, promotes PDAC progression. Careful studies using high density copy number arrays and mutational profiling have excluded the presence of genomic alterations in pancreatic CAFs, potentially suggesting that reprogramming is most likely epigenetic in nature (Walter et al., 2008). While the epigenome of the PDAC neoplastic epithelium has been extensively studied (Sato and Goggins, 2006; Goggins, 2005), the stromal epigenome is largely uncharacterized. Thus, the underlying objective of this study was to study patterns of epigenetic reprogramming in the PDAC stroma, specifically the most predominant α-SMA (ACTA1) expressing CAF cell type (Öhlund et al., 2017).

In the present study, we studied genome wide cytosine methylation in CAFs by using both primary CAFs derived from resected PDAC, as well as de novo CAFs generated from MSCs in vitro. Our analysis revealed widespread loss of DNA methylation in CAFs as the dominant ‘epi-genotype’. This epigenetic reprogramming was associated with upregulation of numerous transcripts, including those encoding the chemokine receptor CXCR4. Our data reveal that stromal CXCR4 overexpression promotes PDAC invasion, and provides a facile druggable target within the tumor microenvironment attenuating tumor progression. Importantly, from a mechanistic standpoint, we determine that paracrine lactate secreted by PDAC cells can be incorporated in stromal cells and lead to increased alpha-keto glutarate (aKG). This is associated with activation of the TET demethylase, thus potentially leading to epigenetic reprogramming seen during CAF formation. Our studies underscore the emerging thread between aberrant metabolism and epigenomic alterations in cancer progression, albeit from the aspect of peritumoral stroma in PDAC.

Results

Widespread epigenetic reprogramming is observed in primary and de novo transformed CAFs

Primary cultures of cancer-associated fibroblasts (CAFs) were established from seven surgically resected PDAC tissue samples and used for epigenomic and transcriptomic analysis. Genome wide cytosine methylation was performed by the HpaII tiny fragment Enrichment by Ligation-mediated PCR (HELP) assay that relies on differential digestion by HpaII and MspI to identify methylated CpG sites (Figueroa et al., 2010a). Unsupervised clustering based on cytosine methylation demonstrated that pancreatic CAFs were epigenetically distinct from other non-cancer associated fibroblast controls that also included hepatic stellate cells. (Figure 1A). To determine the qualitative epigenetic differences between these groups we next performed a supervised analysis of the respective DNA methylation profiles. A volcano plot comparing the differences between mean methylation of individual loci between pancreatic CAFs and non-cancer associated fibroblasts demonstrated that pancreatic CAFs were characterized by widespread hypomethylation when compared to controls (5659 demethylated versus 674 hypermethylated loci in CAFs) (Figure 1B). Gene expression analyses performed on a subset of CAFs also demonstrated transcriptomic differences when compared to controls (Figure 1C). To elucidate the genes that were epigenetically regulated, we analyzed the genes that were concurrently overexpressed and hypomethylated in pancreatic CAFs and observed that critical cellular pathways involved in cell survival, cell cycle and cell signaling were the most significantly deregulated by epigenetically altered genes (Supp File 1). Multiple genes that are known to be important for cell signaling, including secreted interleukins and chemokines such as IL1a, CCL5, CCL26, cellular receptors CXCR4, ICAM3 and signaling proteins MAPK3, MAPK7, JUN were among the easily recognizable genes that exhibited differential hypomethylation and were overexpressed in pancreatic CAFs. Since striking demethylation was observed in primary CAFs, we next wanted to validate these epigenetic changes at a higher resolution in an in vitro model. We generated CAFs from primary mesenchymal stem cells (MSCs) by exposing them to conditioned media from Panc-1 pancreatic cancer (PDAC-CM) cells for 21 days. This method has been shown to transform MSCs into CAFs that are functionally able to support the growth and invasion of malignant cells (Mishra et al., 2008) and resulted in cells with CAF like morphology and higher expression of bona fide CAF markers, aSMA (ACTA1) and FAP (Figure 1D). The methylome of MSCs and de novo CAFs was then studied using the HELP-tagging assay that uses massively parallel sequencing to generate genome wide CpG methylation profiles of >1.8 million loci (Suzuki et al., 2010). We observed that widespread cytosine demethylation was the dominant epigenetic change during transformation of MSCs into de novo CAFs (Figure 1E,F). Loss of methylation upon exposure to PDAC-CM was found to affect all parts of the genome (Figure 1G,H). Both hypo- and hypermethylated differentially methylated regions (DMRs) were distributed in various genomic locations in proportions that were comparable to the distribution of HpaII sites (Figure 1H), thus demonstrating that epigenetic changes were occurring genome-wide during CAF transformation. This is in contrast to absence of genetic alterations in PDAC CAFs, as reported by Walter et al. (2008).

Widespread epigenetic and transcriptomic alterations are seen in pancreatic cancer associated fibroblasts.

(A) Unsupervised clustering of cytosine methylation profiles from seven primary patient-derived pancreatic cancer associated fibroblasts (CAFs) and four healthy controls (Hst: Hepatic stellate cells, Hdf: Human dermal fibroblasts, Hsf: Human skin fibroblasts) shows that CAFs are epigenetically distinct (B) Volcano plot shows that majority of differentially methylated loci in primary patient-derived CAFs are hypomethylated when compared to controls (C) Unsupervised clustering of gene expression profiles shows transcriptomic differences between CAFs and controls (D) Three independent MSCs were exposed to PANC-1 conditioned media (CM) for 21 days and analyzed for α-SMA (ACTA1) and Fibroblast activation protein (FAP) expression. (T-test, p<0.05). (E) Unsupervised clustering based on cytosine methylation shows epigenomic differences between MSCs and de novo generated CAFs. Two independent experiments shown. (F) Volcano plot shows that the majority of differentially methylated loci in in vitro generated CAFs are hypomethylated (G) The differentially methylated regions in de novo generated CAFs are distributed throughout the genome as shown in the circos plot. H:Differentially methylated regions between MSC and de novo generated CAFs are present throughout the genome and mirror the distribution of HpaII sites in the genome. Hypo refers to Hypomethylated DMRs in CAFs. Hyper refers to Hypermethylated DMRs in CAFs when compared to MSCs. Genome refers to distribution of HpaII loci in the genome.

https://doi.org/10.7554/eLife.50663.002

CXCR4 is hypomethylated and overexpressed in pancreatic cancer associated fibroblasts and supports neoplastic cell invasion

Integrative analysis between primary CAFs and de novo CAFs showed a common set of 130 unique promoters that were aberrantly methylated (120 hypomethylated and 10 hypermethylated) (Figure 2A, Venn diagram). This conserved epigenetic signature of CAFs (Supplementary file 2) was able to clearly separate these cells from normal controls and MSCs in supervised clustering (Figure 2A, Bottom panel). The gene promoter encoding for the CXCR4 receptor was found to be hypomethylated in both primary and de novo generated CAFs, and significantly demethylated CpGs were present in CpG ‘shores’, that have been shown to be sites of differential methylation in cancer (Irizarry et al., 2009) (Figure 2B,C). Although recent empirical studies have found upregulation of CXCR4 protein in PDAC cells by immunohistochemical assessment (Bachet et al., 2012), the role of this receptor on the stromal cells have not been studied.

Figure 2 with 1 supplement see all
CXCR4 is demethylated and overexpresed in CAFs and increases pancreatic cancer invasiveness.

(A) A set of 120 transcripts are commonly hypomethylated in primary patient-derived CAFs and de novo generated CAFs and includes CXCR4. (B,C) The CXCR4 promoter is demethylated in primary patient-derived CAFs as seen by the HELP assay (B) and quantitative MassArray Epityper analysis (C). (D - F) CXCR4 knockdown in de novo CAFs leads to abrogation of the increased invasion of Panc1 cells on co-culture. (N = 3, p value<0.05) (G) Co-culture with de novo CAFs leads to increased transwell invasion by Panc-1 cells, that is abrogated after treatment of CAFs with CXCR4 inhibitor AMD-3100 (N = 3, p value<0.05) H: Gene expression profiling of CAFs with CXCR4 knockdown reveals signficantly downregulated (left panel) and upregulated (right panel) transcripts. Salient examples of downregulated transcripts include interleukins IL-8 and IL-33 and the chemokine CCL2.

https://doi.org/10.7554/eLife.50663.003

To determine the functional role of CXCR4 expression on pancreatic CAFs, we used specific siRNAs against CXCR4 that were able to significantly decrease CXCR4 expression in MSC-derived de novo CAFs (Figure 2D, Figure 2—figure supplement 1). Matrigel transwell double chamber invasion assays with PDAC (PANC-1) cells revealed increased invasion of the neoplastic cells in the presence of de novo generated CAF cells (Figure 2E). The increased invasiveness of PDAC cells on co-culture was abrogated with RNAi-mediated knockdown of CXCR4 in the CAFs (t-test, P Value < 0.05) (Figure 2E,F, Figure 2—figure supplement 1). A specific inhibitor of CXCR4, AMD-3100, also led to decreased invasion of PANC-1 cells when cultured with de novo CAFs. (Figure 2G). These were validated in another highly metastatic Pa03C PDAC cell line (Figure 2—figure supplement 1), supporting a role for stromal CXCR4 blockade in attenuating tumor progression within the neoplastic compartment. Finally, gene expression profiling in two sets of CAFs after CXCR4 knockdown revealed significant alterations in the transcriptional profile (Supplementary file 3), including reduced expression of IL8 and other secreted ligands that are known to be tumor microenvironmental regulators (Matsuo et al., 2009; Fang et al., 2016) (Figure 2H).

Tumor mediated lactate flux leads to production of alpha ketoglutarate, TET activation and increased cytosine hydroxymethylation in stromal cells

Next, we wanted to determine whether a diffusible factor secreted by PDAC cells could facilitate the epigenomic reprogramming and demethylation of MSCs to CAFs. Lactate is produced by PDAC cells via lactate dehydrogenase (LDH) enzyme during glycolysis (Le et al., 2010) and has been shown to be an important mediator of metabolic pathways that can regulate the demethylase TET enzymes (Intlekofer et al., 2015; Figueroa et al., 2010b). Thus, we wanted to evaluate whether paracrine lactate, secreted by PDAC cells, could be incorporated by MSCs and result in observed epigenetic changes through modulation of TET activity. Metabolomics analysis using uniformly 13C-labeled lactate in the media as a tracer revealed that primary human MSCs can uptake lactate and convert it to pyruvate and various Krebs cycle intermediates including citrate, alpha-keto glutarate (aKG), fumarate, malate, and aspartate (Figure 3A, Figure 3—figure supplement 1 showing metabolities in parallel experiments using low glucose conditions). We had previously observed a similar metabolic crosstalk between ovarian cancer cells and CAFs, which utilized cancer-secreted lactate as a carbon source (Yang et al., 2016).

Figure 3 with 2 supplements see all
Increased 5hmC is seen in de novo CAFs generated from MSCs and is regulated by a paracrine lactate mediated metabolic flux.

(A) Mapping of carbon atoms transition using U-13C3-lactate. Mass isotopomer distribution (MID) shows uptake of labelled lactate by MSCs cultured in α-MEM leading to the generation of α-KG (M2 αKG) via the Krebs cycle. (B) Conditioned media (CM) from mock treated and LDH inhibitor (FX11) treated Panc-1 (PDAC) cells was collected. MSCs were exposed to media alone, control conditioned media (PDAC CM), or LDH inhibitor treated conditioned media for 14 days in order to generate de novo CAFs (dn-CAF). TET enzymatic activity increases in dn-CAFs after exposure to PDAC CM and is abrogated after exposure to CM from LDH inhibitor treated Panc-1 cells (N = 2, p<0.05). (C) Quantitative analysis of 5-hMC levels by LC-MS demonstrates significant increase within dn-CAFs after treatment with Panc-1 CM, that is abrogated after exposure to CM from LDH inhibitor treated Panc-1 cells (N = 2, p<0.05). (D) CAFs were exposed to exogenous lactate and α-KG levels were analysed by ELISA. (N = 2, p<0.05) (E) 5hmC analysed in the resulting dn-CAFs by LC-MS. After a 2 week exposure to exogenous lactate, significantly increased 5hmC is observed in the dn-CAFs (N = 2, p<0.05). (F) Schematic model of lactate flux from tumor cells to MSCs during CAF differentiation, leading to aKG generation, TET activation and conversion of 5mC into 5hmC.

https://doi.org/10.7554/eLife.50663.005

Since aKG is an important cofactor for TET enzymes, we next determined whether lactate secreted by PDAC-conditioned media could lead to TET activation and increased hydroxymethylcytosine (5hmC) levels in stromal cells. MSCs were exposed to conditioned media from PANC-1 cells treated with an LDH inhibitor or control media for 14 days to induce generation of de novo CAFs. Control PANC-1 conditioned media per se led to significant TET activation (Figure 3B) and increase in 5hmC levels in resulting de novo CAFs (Figure 3C). These effects were, however, not observed when MSCs were exposed to conditioned media from LDH inhibitor-treated PDAC cells. Conditioned media from PDAC cells with LDH knockdown with siRNAs was also able to abrogate the increase in expression of CAF markers (aSMA (ACTA1), FSP and Vimentin) during CAF conversion (Figure 3—figure supplement 2A). Furthermore, exposure of MSCs to exogenous lactate in media was also able to increase aKG levels, TET activity and 5hmC levels after 2 weeks of exposure (Figure 3D,E, Figure 3—figure supplement 2), demonstrating the role of this exogenous metabolite in epigenetic reprogramming of MSCs to de novo CAFs (Proposed Model in Figure 3E). TET activity increased in a dose dependant manner with exogenous lactate and was abrogated by inhibition of mitochondrial pyruvate carrier inhibitor UK5099 (Figure 3—figure supplement 2). Since 2hydroxyglutarate (2HG) and fumarate are inhibitors of TET enzymes, we also observed that exogenous lactate was able to reduce 2HG/aKG and fumarate/aKG ratios in metabolic flux experiments (Figure 3—figure supplement 1). Additionally, exogenous cell permeable aKG was also able to increase TET activity in CAFs (Figure 3—figure supplement 2).

5hmc gains are seen during MSC to CAF conversion

To determine the genes that acquire 5hmC during CAF conversion at a high resolution, we generated CAFs from MSCs exposed to Panc1 conditioned media and used the cells for genome wide 5hmC analysis using Oxidative bisulfite sequencing (OXBS). The 5hmC gains were found to occur throughout the genome (Figure 4A). Genes affected by 5hmC gains were found to group into important regulatory pathways (Figure 4B). Gene associated with 5hmC gains included CAF markers αSMA (ACTA1), FSP1 (S100A4), and Collagen (COL3A1, associated with fibrotic reaction seen in pancreatic cancer) (Figure 4C–E). We also observed 5hmC gain at the CXCR4 promoter as well as downstream enhancer (Figure 4F). The important TGF-β mediator, SMAD2 also acquired 5hmC (Figure 4G) and a transcription factor motif analysis of all sites of 5hmC gain revealed enrichment for smad2 and smad3 binding motifs (Table 1). TGF-beta mediated smad activation is an important regulator of fibrosis and has been shown to be activated in CAFs (Biffi et al., 2019).

5hmC gains are seen during MSC to CAF conversion and occur at gene associated with cancer associated fibroblasts.

(A) Genome wide 5hmC analysis was done by OXBS in MSCs and converted CAFs. 5hmC gains are depicted as percentage gained (CAF/MSC) and were seen to occur throughout the genome (B) 5hmC gains in CAFs were seen to occur at important genetic pathways C-G: 5hmC gains occurred with corresponding decrease in 5mC at genes related to CAF conversion (FSP1 (S100A4), aSMA (ACTA1)); Collagen production (COL3A1), CXCR4, and the SMAD2.

https://doi.org/10.7554/eLife.50663.008
Table 1
Transcription factor sites that are enriched around 5hmC gains during CAF conversion.
https://doi.org/10.7554/eLife.50663.009
RankNameP-valueq-value (Benjamini)# Target Sequences with Motif% of Targets Sequences with Motif# Background Sequences with Motif% of Background Sequences with MotifMotif
1Atf3(bZIP)/GBM-ATF3-ChIP-Seq(GSE33912)/Homer1.00E-0703447.86%2593.15.81%
2AP-1(bZIP)/ThioMac-PU.1-ChIP-Seq(GSE21512)/Homer1.00E-0603838.75%2995.76.71%
3Jun-AP1(bZIP)/K562-cJun-ChIP-Seq(GSE31477)/Homer1.00E-050.00031122.56%722.11.62%
4BMAL1(bHLH)/Liver-Bmal1-ChIP-Seq(GSE39860)/Homer1.00E-050.000385719.57%756716.95%
5Fosl2(bZIP)/3T3L1-Fosl2-ChIP-Seq(GSE56872)/Homer1.00E-050.00031493.40%1040.82.33%
6BATF(bZIP)/Th17-BATF-ChIP-Seq(GSE39756)/Homer1.00E-050.00033227.35%2570.85.76%
7Fra1(bZIP)/BT549-Fra1-ChIP-Seq(GSE46166)/Homer1.00E-040.00062666.07%2089.94.68%
8Smad2(MAD)/ES-SMAD2-ChIP-Seq(GSE29422)/Homer1.00E-040.001165214.89%5725.212.83%
9NPAS2(bHLH)/Liver-NPAS2-ChIP-Seq(GSE39860)/Homer1.00E-040.002654312.40%4734.110.61%
10Smad4(MAD)/ESC-SMAD4-ChIP-Seq(GSE29422)/Homer1.00E-030.00365514.96%5819.313.04%
11Smad3(MAD)/NPC-Smad3-ChIP-Seq(GSE36673)/Homer1.00E-030.0035123828.27%1153325.84%
12Usf2(bHLH)/C2C12-Usf2-ChIP-Seq(GSE36030)/Homer1.00E-030.00422004.57%1574.53.53%
13HIF2a(bHLH)/785_O-HIF2a-ChIP-Seq(GSE34871)/Homer1.00E-030.00661784.06%1393.53.12%
14CEBP:CEBP(bZIP)/MEF-Chop-ChIP-Seq(GSE35681)/Homer1.00E-030.01411072.44%789.31.77%
15MafA(bZIP)/Islet-MafA-ChIP-Seq(GSE30298)/Homer1.00E-020.0233157.19%2709.16.07%
16Bach2(bZIP)/OCILy7-Bach2-ChIP-Seq(GSE44420)/Homer1.00E-020.023861.96%622.71.40%
17Brachyury(T-box)/Mesoendoderm-Brachyury-ChIP-exo(GSE54963)/Homer1.00E-020.03261343.06%1056.72.37%
18USF1(bHLH)/GM12878-Usf1-ChIP-Seq(GSE32465)/Homer1.00E-020.03871904.34%1574.43.53%
19n-Myc(bHLH)/mES-nMyc-ChIP-Seq(GSE11431)/Homer1.00E-020.05112726.21%2354.15.27%
20Fli1(ETS)/CD8-FLI-ChIP-Seq(GSE20898)/Homer1.00E-020.052843910.03%3953.78.86%

Increased stromal 5hmc and CXCR4 expression is seen in in primary human PDAC and murine KPC PDAC tumors

We next wanted to determine the magnitude of increased 5hmC and CXCR4 in stromal cells in large cohort of human primary PDAC samples. 5hmC (Figure 5A) and CXCR4 (Figure 5B) immunohistochemical staining was done on human PDAC TMAs and grading of intensity of stain in the tumor stromal CAFs was estimated. We observed that most CAF like cells in PDAC samples were positive for 5hmC and CXCR4 (1+ to 3+ staining intensity), (Figure 5C). Total PDAC samples examined for 5hmC were 254 and for CXCR4 were 261.

Increased 5hmC and CXCR4 expression is seen in primary human PDAC and murine KPC PDAC tumors.

(A, B) 5hmC (A) and CXCR4 (B) immunohistochemical staining was done on human PDAC TMAs and grading of intensity of stain in the tumor stromal CAFs was estimated. (C) Most CAF like cells in PDAC samples were positive for 5hmC and CXCR4 (1+ to 3+ intensity). Total PDAC samples examined for 5hmC were 254 and for CXCR4 were 261. (D,E) 5hmC (D) and CXCR4 (E) immunohistochemical staining was done on mouse PDAC tumors obtained from 3 KPC mice. CAF like stromal cells were positive for 5hmC and CXCR4. (F) Single cell RNA-seq from low grade intraductal papillary mucinous neoplasm (LG IPMN), high grade intraductal papillary mucinous neoplasm (HG IPMN) and frank pancreatic ductal adenocarcinoma (PDAC) was conducted and shown in tSne plot. (G,H) Stromal cell populations cluster distinctly and are positive for α-SMA (ACTA1) and fibroblast activated protein. (I) Most of the stromal cells are seen in High grade IPMN and PDAC samples J,K: CXCR4 expression is seen in 14/181 (8%) stromal cells and correlates with cells with higher collagen expression, seen in activated CAF phenotypes (T-Test, P Vccal = 0.02).

https://doi.org/10.7554/eLife.50663.010

Next, we wanted to evaluate whether increased stromal 5hmC and CXCR4 was observed in mouse model of PDAC also. PDAC samples from KPC (Kras mutant, tp53 mutant) (Olive et al., 2009) mouse model of were obtained and immunostained for 5hmC and CXCR4. CAF like stromal cells in PDAC tumors from KPC mice were found to be positive for both 5hmC and CXCR4 staining in all tumors examined (Figure 5D,E).

To validate at single cell levels, we next analyzed single cell RNA-seq (scRNAseq) data from samples obtained from precancerous low grade intraductal papillary mucinous neoplasm (LG IPMN), high grade intraductal papillary mucinous neoplasm (HG IPMN) and frank pancreatic ductal adenocarcinoma (PDAC) (Bernard et al., 2019). Stromal cell populations positive for alphaSMA (ACTA1) (Figure 5G) and fibroblast activated protein (Figure 5H) were found to cluster distinctly from malignant cells in tSNE plots. Most of stromal cells were seen in high grade IPMN and PDAC samples (Figure 5I). CXCR4 expression was seen in 14/181 (8%) stromal cells and correlated with cells with higher collagen expression, that is seen in activated CAF phenotypes (Öhlund et al., 2017) (TTEst, P Val = 0.02)(Figure 5K).

TET-deficient MSCs lead to inefficient CAF conversion and reduced tumor growth in vivo

Having demonstrated that exposure to PDAC-conditioned media lead to TET activation with a concomitant increase in 5hmC levels within CAFs, we next wanted to determine the functional role of TET enzymes during MSC to de novo CAF differentiation. MSCs were obtained from TET2 KO mice (Dawlaty et al., 2014) and controls and were co-cultured with conditioned media from murine PDAC cells derived from Kras, p53 mutant tumors (‘KPC’ cells) (Torres et al., 2013). WT MSCs acquired fibroblastic ‘CAF-like’ appearance after exposure to PDAC conditioned media, while TET2 KO MSCs generally retained their original morphology (Figure 6A). Additionally, TET2 KO MSCs that were exposed to CM led to significantly less KPC PDAC cell invasion in matrigel when compared to WT controls (Figure 6B,C). Next, to determine the functional role of TET demethylase in CAF generation in vivo, we co-injected murine KPC cells with MSCs from TET2 KO mice and controls into immune-deficient mice. Co-injected murine PDAC cells with TET2KO MSCs had significantly slower growth rates in vivo when compared to controls (Figure 6D). Explanted allografts with TET2 KO MSCs were significantly smaller (Figure 6E,F) and histologically revealed significantly less cells with a CAF phenotype, on staining with aSMA (ACTA1) (Figure 6G,H), thus demonstrating less efficient CAF conversion and tumor supporting capabilities in vivo.

TET2 inactivation in MSCs leads to reduced tumor growth and inefficient CAF conversion in vivo.

(A) Wildtype and TET2 KO MSCs were exposed to conditioned media from KPC PDAC cells for 14 days. TET2 KO MSCs demonstrated decreased fibroblastic appearance when compared to wildtype controls. (B,C) MSCs from TET2 KO treated with conditioned media from KPC PDAC cells led to significantly less KPC invasion in matrigel assays when compared with wildtype controls (N = 3 independent experiments, p value<0.05) (D) MSCs from TET2 KO mice and WT controls were co-injected subcutaneously with KPC PDAC cells in NSG mice. Tumor growth was significantly reduced in mice injected with TET2 KO stromal cells (N = 9 (TET2 KO/KPC) and N = 7 (WT/KPC), P Value < 0.05) (E,F) Tumor sizes (B) and weights (C) are significantly reduced from KPC allografts co-injected with TET2KO MSCs (P Value < 0.05) G,H: Number of aSMA (ACTA1) positive cancer associated fibroblasts are reduced in KPC allografts co-injected with TET2KO stromal cells when compared to KPC allografts co-injected with control MSCs (N = 4 in each group, p value<0.01).

https://doi.org/10.7554/eLife.50663.011

Discussion

The tumor microenvironment plays a critical role in promoting the growth and invasion of cancer cells (Karnoub et al., 2007; Orimo and Weinberg, 2006). Cancer cells can recruit MSCs from the marrow and facilitate their transformation into activated CAFs, through a plethora of paracrine signals, such as chemokines (Mishra et al., 2011; Quante et al., 2011). One aspect of the cancer cell ‘secretome’ within the immediate juxtatumoral milieu that has not been fully examined pertains to the role of secreted metabolites, or metabolic intermediates. We demonstrate that conversion of MSCs into CAFs is associated with widespread epigenomic reprogramming. Specifically, we establish that tumor generated lactate can be incorporated by stromal cells and can potentially induce epigenomic changes via increased production of aKG. Notably, aKG is an essential cofactor for the functionality of TET demethylates (Carey et al., 2015). Studies in embryonic stem cells demonstrated the ability of lactate in reprogramming stem cells via epigenetic alterations (Carey et al., 2015). It has also been shown that tumor lactate can cause pleiotropic effects in surrounding immune cells, as well as in the tumor cells themselves (Matilainen et al., 2017). Our data demonstrate that this diffusible factor can be a potential critical mediator that facilitates CAF differentiation in the vicinity of glycolytic proliferative tumors such as PDAC. Lactate inhibitors are being developed for anti-tumor activity (Le et al., 2010; Rajeshkumar et al., 2015) and our data suggests that these inhibitors may act via effects on tumor microenvironment also.

Epigenetic reprogramming, which manifests as widespread loss of DNA methylation and gain of cytosine hydroxymethylation at selective promoters, is seen in both MSC-derived and primary (patient-derived) pancreatic CAFs. Increased lactic acid can result in acidic environment and it is possible that the change in pH can also influence epigenetic states. Previous studies have shown that a low acidic ph leads to increases in 2HG more than aKG (Nadtochiy et al., 2016). 2HG increase generally leads to decreased Tet activity. Thus in our model, a low ph (due to increase lactic acid) should not account for the high Tet activity and consequent decreased 5mC that we see; suggesting that lactic acid mediated increases in aKG and Tet activity were not influenced by changes in local pH. Loss of DNA methylation has been described mainly during developmental processes of early embryo development, and also during differentiation of hematopoietic stem cells to committed red cell progenitors (Shearstone et al., 2011; Yu et al., 2013). In fact, involvement of a large proportion of the genome by demethylation is rarely outside of developmental processes. Also, though hypomethylation has been shown in some solid tumors (Timp and Feinberg, 2013; Alvarez et al., 2011), it has not been studied at single base pair resolution in the tumor microenvironment. Epigenetic studies in the tumor microenvironment have mainly been single locus studies that have focused on hypermethylation of specific gene promoters during CAF transformation. Our findings show that widespread demethylation occurs during in vitro transformation of MSCs to CAFs, and is recapitulated in primary patient-derived CAF samples. In fact, genome wide analysis of 5hmC showed that 5hmC acquisition was seen in genes that have been associated with pancreatic cancer associated CAFS. 5hmC is an epigenetic modification that is obtained from oxidation of 5mC marks and is an intermediary step towards demethylation (Ko et al., 2010). It is postulated that 5hmC can act as an independent regulatory activating mark and is associated with sites of active transcription (Bhattacharyya et al., 2017; Madzo et al., 2014). Furthermore, our findings are consistent with a study in gastric cancer that also observed loss of methylation and was consistent with our findings. (Jiang et al., 2008) A recent immunohistochemical study in a murine model of PDAC also observed loss of methylation in the microenvironment (Shakya et al., 2013) though it did not study locus specific changes.

Our data shows that the chemokine receptor, CXCR4, is upregulated in CAFs and is associated with loss of promoter methylation. Interestingly the promoter CpGs that are demethylated were not located in CpG islands, but in the neighboring CpG shore. CpG shores flank CpG islands and have been shown to areas where differential methylation can occur in cancer (Irizarry et al., 2009). CXCR4 is a well-studied receptor in stem cells and cancer models (Rettig et al., 2012; Wong and Korz, 2008) and most studies have evaluated its expression on the tumor cells. We show that in addition to its roles on tumor cells, its expression on stromal cells is also functionally important in regulating tumor cell invasion. Since host stromal cells and the cancer cells cross-talk via a large variety of soluble factors, chemokines are important mediators of these processes. It has been shown that CAFs (Mishra et al., 2008) as well as tumor cells (Quante et al., 2011) secrete SDF-1/CXCL12, which is the ligand for CXCR4. A recent study demonstrated that overexpression of the CXCR4 ligand, SDF-1 in gastric cells can induce myofibroblast expansion and is consistent with the role of CXCR4 in pancreatic cancer (Quante et al., 2011; Shibata et al., 2013). The specific inhibitor of CXCR4, AMD-3100, is a FDA approved drug for hematopoietic stem cell mobilization. Our results demonstrate a role of this drug in inhibition of tumor stromal crosstalk. We also observed that production of secretory factors, such as interleukin-8 (IL8) by CAFs, is reduced upon CXCR4 knockdown. IL8 is a well-known regulator of cell motility and has been shown to be a regulator of pancreatic cancer cell invasion and growth (Matsuo et al., 2009; Delitto et al., 2017; Sparmann and Bar-Sagi, 2004).

Numerous studies have shown that depletion of stroma in PDAC can lead to enhanced efficacy of chemotherapy (Feig et al., 2012; Öhlund et al., 2017; von Ahrens et al., 2017; Ireland et al., 2016). Conversely, a murine study demonstrated that depletion of CAFs can accelerate PDAC metastases (Özdemir et al., 2014). As the prognosis of PDCA remains dismal, various clinical approaches are being attempted to target CAFs, and will further clarify the role of these cells in carcinogenesis. Our studies support the use of agents that disrupt the cross talk between malignant cells and the stroma and suggest that CXCR4 inhibitors may have a potential therapeutic role in pancreatic cancer that should be tested in future studies.

Materials and methods

DNA methylation analysis by HELP assay

Request a detailed protocol

Genomic DNA was isolated from primary CAF samples and controls with the use of a standard high-salt procedure, and the HELP assay was carried out as previously described (Alvarez et al., 2011; Khulan et al., 2006). The assay uses comparative isoschizomer profiling, interrogating cytosine methylation status on a genomic scale. Briefly, genomic DNA from the samples was digested by a methylcytosine-sensitive enzyme HpaII in parallel with MspI, which is resistant to DNA methylation, and then the HpaII and MspI products were amplified by ligation-mediated PCR. Both amplified fractions were submitted to Roche-NimbleGen, Inc (Madison, WI) for labeling and hybridization onto a human hg17 custom-designed oligonucleotide array (50-mers) covering 25,626 HpaII amplifiable fragments (HAF) located at gene promoters. HpaII amplifiable fragments are defined as genomic sequences contained between two flanking HpaII sites found within 200–2,000 bp from each other. Each fragment on the array is represented by 15 individual probes distributed randomly spatially across the microarray slide. Thus the microarray covers 50,000 CpGs corresponding to 14,000 gene promoters. Signal intensities at each HpaII amplifiable fragment were calculated as a robust (25% trimmed) mean of their component probe-level signal intensities after intensive quality control using analytical pipelines. The log2(HpaII/MspI) was used as a representative for methylation and analyzed as a continuous variable. For most loci, each fragment was categorized as either methylated, if the centered log HpaII/MspI ratio was less than zero, or hypomethylated if on the other hand the log ratio was greater than zero.

HELP-tagging for genome-wide methylation analysis of in vitro generated CAFs

Request a detailed protocol

The HELP-tagging assay applies massively parallel sequencing to analyze the status of 1.8 million CpGs distributed across the entire genome (Suzuki et al., 2010; Bhattacharyya et al., 2013). To perform HELP-tagging assays, DNA samples were digested with HpaII and ligated to customized Illumina adapters with a complementary cohesive end. These adapters also contain an EcoP15I site that cuts into the adjacent sequence 27 bp away, allowing us to polish that end and ligate the other Illumina adapter for library generation by PCR. The presence of the CCGG and EcoP15I sequences at the ends of the reads allowed us to remove spurious sequences. Prior to sequencing, we performed qRT-PCR with primers that measure the proportion of adapter dimer complexes in the library, usually a very small proportion (<5%) of the total library. Following sequencing, we removed low quality or unmapped reads, piled up reads on each locus and created an output for each locus in terms of read frequency. We normalized the HpaII signal with that of the deeply-sequenced MspI profiles, as performed previously (Suzuki et al., 2010; Wu et al., 2013). Results were generated using the WASP system and linked to a local mirror of the UCSC Genome Browser for visualization. HELP-tagging data were analyzed using an automated pipeline, as described previously (Suzuki et al., 2010; Wu et al., 2013). Loci were defined in a continuous variable model, given the quantitative nature of this and comparable published assays (Ball et al., 2009). Methylation values were depicted from a range of 0 to 100, with 0 representing fully methylated to 100 representing fully hypomethylated loci.

Whole-genome hydroxymethylation analysis by oxidative bisulfite sequencing

Request a detailed protocol

One microgram genomic DNA from MSC and CAF cells were sonicated to 100–400 bp by Bioruptor, and 0.5% (w/w) sequencing spike-in control DNA was added thereafter and purified by Ampure XP beads. Spike-in controls were added to the adapted library. Half of the library was subjected to oxidation reaction following the manufacturer's protocol (Cambridge Epigenetix). Both oxidized and nonoxidized samples were then treated with bisulphite conversion reagent. The final PCR was performed according to the manufacturer's guide using 10 cycles of amplification, purified, and sequenced at the Einstein Epigenomics Facility. Bismark was used to map the sample reads and make methylation calls. At every base location, the 5-mC percentage was estimated by the ratio of nonconverted CpG bases to the total number of bases using the pileup of reads from the OxBS sample. For the estimation of 5-hmC, both BS and OxBS samples were analyzed, and an estimate for the percentage of 5-hmC methylation was calculated by the difference between BS and OXBS conversions.

Since 5-hmC is a less frequent modification, for further stringency in measuring the difference in ratios, we used Fisher's exact test of proportions, using the number of converted and nonconverted reads in the BS and OxBS samples, and selected sites that have a p-value<0.05. 5-mC sites were calculated by the ratio of nonconverted bases to total bases in the OxBS sample with a biologically influenced threshold of 50%.

x.hmc.caf=counts.msccaf[caf.hmc>msc.hmc and caf.hmc >= 0.75 and caf.mc <msc.mc,]

x.hmc.msc=counts.msccaf[caf.hmc<msc.hmc and msc.hmc >= 0.75 and caf.mc >msc.mc,]

CAF hits are the loci where CAF has a higher HMC score than MSC, has an HMC score of at least. 75, and has a lower MC score than MSC.

MSC hits are the loci that have a higher hmc score of at least. 75, and a lower MC score.

To compare hydroxymethylation between cancer and control samples, we used Fisher's test and adjusted for multiple comparisons through the Benjamini-Hochberg procedure.

Quantitative DNA methylation analysis by MassArray epityping

Validation of HELP microarray findings was carried out by MALDI-TOF mass spectrometry using EpiTyper by MassArray (Sequenom, CA) on bisulfite-converted DNA, as previously described (Alvarez et al., 2011; Figueroa et al., 2008; Figueroa et al., 2009). MassArray primers were designed to cover the flanking HpaII sites for a given locus, as well as any other HpaII sites found up to 2,000 bp upstream of the downstream site and up to 2,000 bp downstream of the upstream site, in order to cover all possible alternative sites of digestion.

Gene expression microarrays

Request a detailed protocol

Gene expression data were obtained using Affymetrix Human Genome U133A 2.0 or Plus2 GeneChips; mRNA isolation, labeling, hybridization, and quality control were carried out as described before (Alvarez et al., 2011). Raw data were processed using the Robust Multi-Averaging (RMA) algorithm and Affymetrix Expression Console software. Data are available in the NCBI Gene Expression Omnibus database (GSE101082).

Microarray data analysis

Request a detailed protocol

Unsupervised clustering of HELP and gene expression data by principal component analysis was performed with the use of R 2.8.2 statistical software. Supervised analysis of the methylation data were carried out with a moderated t-test with Benjamini-Hochberg correction with a significance level of P less than. 05 and an absolute difference in methylation greater than 1.5 between the means of the two populations (eg, MSCs vs. CAFs) to increase the likelihood of detecting biologically significant changes in methylation levels.

Gene network and gene ontology analysis

Request a detailed protocol

Ingenuity Pathway Analysis software and the Database for Annotation, Visualization and Integrated Discovery were used to carry out network composition analyses.

In vitro generation of CAFs

Request a detailed protocol

PANC-1 cells were obtained from ATCC (verified by STR authentication) and were grown in DMEM (Life technologies) + 10% heat-inactivated FBS culture medium and were mycoplasma free. Conditioned medium from pancreatic cancer cell conditioned media (CM) was harvested after 16 hr and centrifuged at 3,000 rpm for 5 min and supernatant was passed through Millipore sterile 50 mL filtration system with 0.45 μm polyvinylidene difluoride membrane. Human mesenchymal stem cells (hMSCs) were exposed to fresh pancreatic cancer cell CM continuously, with the medium changed every third day for the entire 21 day time period.

Cell lines and reagents

Request a detailed protocol

Primary cultures of cancer associated fibroblasts (CAFs) were established from excess tissues of surgically resected pancreatic cancers at the Johns Hopkins Hospital. The excess tissues were obtained delinked from direct patient identifiers, and primary (i.e., non-immortalized) CAFs established by passaging in vitro, as previously described (Walter et al., 2010). All CAFs were used at early passage numbers (passages 3–6), and absence of neoplastic epithelium was confirmed by absence of cytokeratin 19 transcripts by qRT-PCR (data not shown). For controls, dermal fibroblasts (Hdf), human skin fibroblasts (Hsf) and hepatic stellate (Hst) cells were obtained from ATCC. The human telomerase reverse transcriptase (hTERT) immortalized CAF line, CAF19, was a kind gift from Dr. Michael Goggins at Johns Hopkins University (Yu et al., 2012). The human pancreatic ductal adenocarcinoma cell line Pa03C (Jones et al., 2009), generated from a liver metastasis, was maintained in DMEM complete media supplemented with 10% FBS and 1% penicillin-streptomycin under mycoplasma free conditions. Mouse pancreatic cancer KPC cells were obtained from Dr Batra (Torres et al., 2013). LDH inhibitor FX11 was obtained from Calbiochem and used as previously (Le et al., 2010).

RNA extraction and quantitative real-time reverse transcription PCR

Request a detailed protocol

Total RNA was extracted using RNAeasy Mini Kit (Qiagen, Valencia, CA). RNA was reverse transcribed using the TaqMan One-Step RT-PCR Master Mix Reagents Kit (Applied Biosystems, Foster City, CA). Quantitative RT-PCR was carried out using a pre-designed gene expression assay for CXCR4 (Applied Biosystems) on a StepOnePlus Real-Time PCR System. Relative fold expression was determined and normalized to GAPDH (Applied Biosytems) levels using the 2(-ΔΔCT) method (Schmittgen and Livak, 2008).

Transfection of small interfering RNA (siRNA)

Request a detailed protocol

De novo (i.e., MSC-derived) CAFs and the immortalized CAF19 cells were seeded and allowed to adhere overnight in 6-well culture plates at a density of 2 × 105 cells/well. Following overnight incubation, cells were transfected with 50 nmol/L siRNA targeting CXCR4 (Dharmacon Technologies, Thermo Fisher Scientific, Lafayette, CO) or non-targeting control siRNA (Dharmacon Technologies) using DharmaFECT four transfection reagent. At 24 hr post-transfection, cells were plated for co-culture invasion assays.

Transwell coculture invasion assay

Request a detailed protocol

8 µm pore size inserts were coated with 100 µL Matrigel (1:40 Matrigel: PBS solution) (BD Biosciences, San Jose, CA) and allowed to solidify in a notched 24-well culture plate overnight. De novo CAFs or CAF19 cells were then plated in the bottom chamber at a density of 5 × 104 cells/well and allowed to adhere overnight. The media was then replaced with DMEM containing 1% FBS and Pa03C or PANC-1 cells were suspended in DMEM containing 0.5% FBS and seeded at 5 × 104/well in the top (notched insert) chamber. Following 48 hr incubation, the assay was terminated and cells migrating to the underside of the insert were fixed in ethanol and stained with 0.25% crystal violet solution. Each condition was performed in triplicate. Invasion assays were also performed in the presence of the CXCR4 antagonist, AMD3100 (Sigma Aldrich, St. Louis, MO), which was added to the lower chamber.

Analysis of global DNA methylation and hydroxymethylation by mass spectrophotometer

Request a detailed protocol

Genomic DNA was hydrolyzed by DNA Degradase Plus (Zymo Research, CA, USA) according to the manufacturer’s instructions. Digested DNA was injected onto a UPLC Zorbax Eclipse Plus C18 RRHD column (Agilent Technologies, CA). The analytes were separated by gradient elution using 5% methanol/0.1% formic acid (mobile phase A) and 100% methanol (mobile phase B) at a flow rate of 0.25 ml/min. Mobile phase B was increased from 0% to 3% in 5 min, to 80% in 0.5 min, kept at 80% for 2 min then switched to initial conditions in 2.5 min. The effluent from the column was directed to the Agilent 6490 Triple Quadrupole mass spectrometer (Agilent Technologies, CA). The following transitions were monitored: m/z 228.1 - > 112.1 (C); m/z 242.1- > 126.1 (5mC) and m/z 258.1- > 142.1 (5hmC).

Calibration solutions with varying amounts of 5hmC (0–3%), 5mC (0–10%) and fixed amount of C, were also analyzed together with the samples. The solutions were prepared from a 200 bp DNA standards containing 57 cytosines which are homogeneous for C, 5hmC or 5mC. Calibration plots of % 5hmC or %5mC vs MRM Response ratio were constructed based on the data obtained. %5hmC is obtained from the ratio of [5hmC]/5hmC]+[C]. Response ratio is the response peak area for 5hmC or 5mC divided by the combined peak areas of 5hmC, 5mC and C. The % 5hmC or 5mC in the samples were determined from the calibration plots.

TET enzymatic activity

Request a detailed protocol

Tet activity in nuclear cell lysates was assessed by Epigenase 5mC-Hydroxylase TET Activity/Inhibition Assay Kit (Colorimetric, Epigentek). The kit contains cofactors needed for Tet activity in vitro (Ascorbic Acid, aKG and FeNH4SO4) and assesses the amount of active TET enzyme in cell lysates based on efficiency of conversion of 5mC to 5hmC.

Immunohistochemistry

Request a detailed protocol

Tissues/cells were fixed in 10% buffered formalin, embedded in paraffin and sectioned using a microtome and mounted onto glass slides. The slides were incubated at 60 C for an hour to melt the paraffin, followed by dehydrating them through gradients of ethanol (70, 80, 90% and 100%) and 100% xylene. The samples were then treated with antigen unmasking solution (Dako pharma) followed by permeabilization with 0.3% H2O2 and blocked using blocking buffer (5% donkey serum and 2% BSA). Samples were then incubated overnight in the primary antibody prepared in the blocking buffer followed an incubation with appropriate HRP conjugated secondary antibody. Color development was achieved by treating the samples with diaminobenzidine (DAB) and counterstaining performed using harris hematoxylene (Dako pharma). The samples were then passaged through alcohol grades and xylene to dehydrate them, mounted using permount solution (fisher scientific) and allowed to dry overnight before the image analysis.

In vivo experiments with TET KO mouse stroma coinjection

Request a detailed protocol

TET2 Knockout (KO) and wild type (WT) C57B6 mice were euthanized according to protocol and their femur bones were harvested. The bone marrow was flushed and the resulting cells were grown to 30–40% cell density. Once the cells were adhered to the culture flask surface, the WT and KO storm cells were divided into two groups. One group was treated with media conditioned with the KPC cells and the control group was treated with plain culture medium every other day for two weeks. The resulting cells were used for downstream experiments. KPC cells (5 million/mouse) were injected along with TET2 KO stromal or WT stromal cells (1 million/mouse) into NOD-scid IL2Rgamma null (NSG) immuno-deficient mice and then followed for tumor measurements. Mice were sacrificed at end of experiment and tumors used for immunohistochemistry.

Stable metabolite isotope analysis using GC-MS

Request a detailed protocol

Metabolic extraction MSCs were cultured with 10 mM 13C3 lactate (Cambridge Isotope Labs) for 48 hr. Spent medium was removed, and cells were washed with ice-cold PBS. Cells were then quenched with 400 µl methanol and 400 µl water containing 1 µg norvaline. Cells were scraped, washed with 800 µl ice-cold choloform, vortexed at 4°C for 30 min and centrifuged at 7,300 rpm for 10 min at 4°C. The aqueous portion was then collected and stored at −80°C until further analysis.

Derivatization Samples were first dried and dissolved in 30 µl of 2% methoxyamine hydrochloride in pyridine (Pierce) prior to sonication for 15 min. Samples were then kept at 37°C for 2 hr, and transferred to 55°C for 1 hr following the addition of 45 μl MBTSTFA+1% TBDMCS (Pierce).

GC-MS measurements Analysis was done using an Agilent 6890 GC equipped with a 30 m Rtc-5 capillary column connected to an Agilent 5975B MS operating under electron impact ionization at 70 eV. Samples were injected at 1 µl and 270°C in splitless mode, and helium was used as the carrier gas at 1 ml/min. The heating cycle for the GC oven was as follows: 100°C for 3 min, followed by 300°C at 5°C/min temperature increase, for a total run time of 48 min per sample. Data was acquired in scan mode and the integrated signal of all potentially labeled ions was normalized by the norvaline signal and used to calculate the abundance of relative metabolites. The mass isotopomer distribution was obtained by dividing the signal of each isotopomer by the sum of all isotopomer signals and corrected for natural abundance. These stable isotope tracer analysis protocols are established and described in detail in our previous studies (Achreja et al., 2017; Zhao et al., 2018).

Nuclear protein extraction and In vitro TET enzymatic activity analysis

Request a detailed protocol

Cells were treated with various conditions and nuclear protein was then isolated from cells using the EpiQuik nuclear extraction kit (Epigentek Group Inc), according to the manufacturer’s instructions. TET enzymatic activity was measured by using the ELISA-based Epigenase 5mC Hydroxylase TET Activity/Inhibition Assay Kit (Fluorometric) according to manufacturer's instructions. This technique relies on the conversion of methylated products at the bottom of the wells to hydroxymethylated products by the TET enzyme present in the nuclear extract. Thus the amount of hydroxymethylated products formed is a measure of the TET activity of the nuclear extract harvested from the cells being tested.

Data availability

Sequencing data have been deposited in GEO under accession code GSE135218.

The following data sets were generated
    1. Verma A
    2. Pradhan K
    (2019) NCBI Gene Expression Omnibus
    ID GSE135218. Lactate-mediated Epigenetic Reprogramming Regulates Formation of Human Pancreatic Cancer-associated Fibroblasts.

References

Decision letter

  1. Ralph DeBerardinis
    Reviewing Editor; UT Southwestern Medical Center, United States
  2. Anna Akhmanova
    Senior Editor; Utrecht University, Netherlands

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The paper describes widespread epigenetic changes, particularly broadly reduced cytosine methylation, in cancer associated fibroblasts from pancreatic ductal adenocarcinoma (PDAC). These changes are caused by enhanced activity of TET methylcytosine hydroxylases, and result in enhanced expression of CXCR4, which promotes cancer cell invasiveness. Metabolic experiments indicate that cultured PDAC cells secrete lactate, which is taken up by cultured CAFs and used to maintain a large intracellular alpha-ketoglutarate pool, which promotes TET activity. The authors conclude that cancer cell lactate release has a paracrine effect to reprogram gene expression in stromal cells. With increasing appreciation for lactate's involvement in tumor metabolism beyond its presence as a by-product of excessive glycolysis, these new findings are timely and will be interesting to investigators in cancer metabolism and the tumor microenvironment.

Decision letter after peer review:

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Paracrine Lactate Mediated Epigenetic Reprogramming Regulates Formation of Pancreatic Cancer-associated Fibroblasts" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife. Although the reviewers found some aspects of the study interesting, there were concerns about conceptual novelty and methodological rigor that will make it difficult to revise the manuscript within the usual 2 month time frame recommended for revision of eLife papers. Given that the reviewers found the area of research to be interesting, should you be able to thoroughly address the critiques, the editorial team would consider a revised paper as a new submission.

Reviewer #1:

These authors studied the role of epigenetic reprogramming in cancer-associated fibroblasts (CAFs) from pancreatic ductal adenocarcinoma (PDAC). They find near-global DNA hypomethylation both in primary CAFs and in CAFs generated 'de novo' from mesenchymal stem cells cultivated in medium conditioned by PDAC cells. Hypomethylation was associated with widespread changes in gene expression, including the induction of numerous signaling genes. One such hypomethylated/overexpressed gene in CAFs was CXCR4, and siRNA silencing of this gene in CAFs was sufficient to suppress PDAC cell invasiveness. The authors then use 13C-base isotope tracing to show that lactate, a product secreted by PDAC cells, can be taken up by CAFs and used to supply the TCA cycle. 13C-lactate contributes to part of the CAF aKG pool, and this appeared to be associated with function of TET, an aKG-dependent enzyme, because inhibiting LDH suppressed TET activity. TET inactivation in MSCs also led to reduced production of CAFs and reduced xenograft growth in vivo. The authors conclude that lactate metabolism enables epigenetic reprogramming in CAFs, and that this in turn supports PDAC cell invasion and growth.

This area of work is potentially interesting because relatively little is known about metabolic cross-talk between cell types within the tumor microenvironment. The finding that TET is required for the MSC-to-CAF conversion is also interesting. However, there are several significant conceptual and methodological flaws with the paper in its current form.

1) In Figure 1, epigenetic marks are compared between CAFs and control stromal cells. This is problematic because the CAFs were generated from primary tumor samples and used at low passage, while the controls were purchased from ATCC and no information about passage number is provided. Given the known, potentially widespread epigenetic changes that occur in culture, the large changes reported in the figure could be due to culture-related artifacts. The type of culture medium, which is not described in detail for the cell lines used in the experiment, can also influence these marks. Overall, it is not clear that the authors are studying relevant changes in fibroblast biology.

2) Along these lines, with some 75% of the genome displaying altered cytosine methylation in CAFs, there is remarkably little overlap between promoters affected in primary and de novo-produced CAFs (Figure 2). This raises concern that the de novo CAFs are not particularly good models with which to study CAF epigenetics.

3) In the experiments to manipulate CXCR4, only a single siRNA is used, silencing of mRNA levels does not appear very effective (only about 50% silencing in Figure 2D), no western blots are shown to document reduction in protein levels, and no attempts are made to reverse the effects (e.g. by expressing a non-silenced allele of CXCR4), making it hard to conclude that the siRNA's suppression of PDAC cell invasion are on target.

4) In the metabolism experiments, the authors provide a high dose of 13C-lactate (10 mM) and culture the cells for 48 hours. Yet only 10% of the aKG pool is labeled. This observation provides strong evidence against the main conclusion of the paper: that lactate is an important source of aKG. Citrate is much more extensively labeled by lactate (40%), suggesting that most aKG arises from another carbon source, probably glutamine.

5) The authors should add lactate to the PDAC-conditioned medium in the LDH inhibitor experiment. This would provide evidence that the inhibitor's effect was on target. It would also be useful to perform this experiment genetically by silencing or mutating LDH in the CAFs – this should prevent lactate from being used as a carbon source in these cells.

6) Do the CAFs have any particular predilection to use lactate relative to the PDAC cells? This is an important question, because in the tumor microenvironment, these cells could presumably be competing for lactate.

Reviewer #2:

The present manuscript examines the links between the classical tumor cell metabolite lactate and CAF epigenetics in PDAC. The paper also zooms in on CXCR4 and TET2 as particular genes whose expression in CAFs is important for PDAC growth. The paper also proposes a mechanism by which lactate controls TET2 via KG. Overall, this paper examines a trendy idea (tumor cell secreted products impacting CAF epigenetics) and is concise and an appealing read, but I have serious concerns about the novelty, the quality, and especially the validity of the mechanism:

1) Figure 3 shows that labeled lactate labels TCA intermediates. This is something that was apparent to metabolism experts for ~ 50 years and shown beyond a shadow of a doubt by the Nat Chem paper from Patti lab.

2) Labeling of KG is small (since most KG comes from glutamine in this [not fully physiologic] media). It is most unlikely that 10% of KG coming from lactate can have any affect on epigenetics. There is no evidence presented for altered KG concentration. It is the concentration that controls TET2 activity.

3) The TET2 activity assay is in some version of lysates, which seemingly should not depend on cellular KG as the substrate (methods are not very complete, just referencing a kit). There is no clear logic connecting these biochemical TET2 measurements to KG or lactate. (Obviously lactate can impact cells many ways: pH, energy, pyruvate, NADH, etc.)

4) P-values are reported for inappropriately small N (even N = 2).

5) It is hard to find concentrations of small molecule inhibitors employed, and there is no dose titration for lactate (used at 10 mM-- higher than reported values in PDAC tumors) or for small molecule tool compounds.

6) siRNA KD seem to be reported for single siRNA (evidence is that 5+ are needed for confirming target selectivity given frequent off-target effects) and without rescue (which is of great importance for proving on-target mechanism); there is no use of more advanced and reliable genetics.

7) There are no experiments actually testing paracrine lactate-mediated epigenetic reprogramming in a physiological context.

In summary, the paper is nicely constructed but limited in conceptual novelty and does not come close to proving the proposed mechanism from lactate-->KG-->Tet-->hypomethylation-->CXCR4 expression.

Reviewer #3:

This is an excellent study highlighting the role of metabolism reprograming on epigenetics of stromal cells to control tumor growth. I have a few experiments to bolster their conclusions.

Experiments:

1) The authors suggest that lactate is a fuel to support TCA cycle metabolite generation. I assume lactate becomes pyruvate that enters the mitochondria where it is converted into acetyl-CoA via PDH. The authors should prevent pyruvate entry into the mitochondria by using mitochondrial pyruvate carrier inhibitor UK5099 and examine methylation status.

2) Does cell permeable aKG recapitulate the effects of lactate (2 weeks) on methylation in CAFs?

3) It is possible that the presence of lactate diminishes glycolytic flux, which could diminish one-carbon metabolism (SAM levels). They should measure SAM levels as well as provide ratios of succinate/aKG, 2HG/aKG and fumarate/aKG, which have been shown to inhibit TETs, in CAFs +/-lactate.

4) Can they recapitulate the KPC + TET2 KO CAF tumor experiments with KPC + WT CAFs treated with lactate for 2 weeks (Figure 4G)?

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for submitting your article "Lactate-mediated epigenetic reprogramming regulates formation of pancreatic cancer-associated fibroblasts" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Anna Akhmanova as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The paper describes widespread epigenetic changes, particularly broadly reduced cytosine methylation, in cancer associated fibroblasts from pancreatic ductal adenocarcinoma (PDAC). These changes are caused by enhanced activity of TET methylcytosine hydroxylases, and result in enhanced expression of CXCR4, which promotes cancer cell invasiveness. Metabolic experiments indicate that cultured PDAC cells secrete lactate, which is taken up by cultured CAFs and used to maintain a large intracellular alpha-ketoglutarate pool, which promotes TET activity. The authors conclude that cancer cell lactate release has a paracrine effect to reprogram gene expression in stromal cells.

Essential revisions:

1) There needs to be more clarity about the assays used to measure TET activity in lysates. The authors are proposing that aKG abundance from the lysate influences the activity reported by this commercial assay. But does the assay buffer contain aKG? If so, it seems unlikely that the relatively small amount of aKG in the lysates could substantially affect TET activity. Please clarify: (i) The source and abundance of aKG in the assay, which is not clear from the methods or the vendor website; and (ii) based on the answer, a clear delineation in a few sentences of reasonable possibilities of how components in the lysate (including aKG) might influence the results of the assay.

2) In the original submission, two reviewers questioned the importance of lactate for formation of aKG when isotope tracing with 13C-lactate resulted only in very low labeling of the aKG pool. The explanation for this result seems to be that lactate-derived synthesis of alanine results in production of unlabeled aKG; therefore observation of labeled aKG underestimates lactate's contribution. If this is the argument, it should be spelled out clearly in the manuscript itself, because the impression is still that lactate makes only a small contribution to aKG. The authors also need to explain why the fractional enrichment in alanine is so different between Figure 3A and Figure 3—figure supplement 1A. The latter was used to make the point about alanine labeling in the rebuttal, but the enrichment in both high and low glucose conditions in this experiment is much higher than in the main figure. Finally, it would help the reader if the "accessory" formation of aKG via alanine synthesis (i.e. that pyruvate's conversion to alanine results in glutamate's conversion to aKG) was clearly illustrated in the model in Figure 3F.

3) Evidence for paracrine lactate exchanges in vivo is still circumstantial. Some of the same phenomena are observed in culture and in vivo, and the evidence for TET2's involvement in vivo is strong. But there is still no compelling evidence that lactate released by cancer cells has anything to do with TET2 activity in CAFs in the tumor microenvironment. Adding direct evidence along these lines would strengthen the paper. Failing that, the authors need to be more circumspect about the role of this mechanism in tumors.

4) The authors should address how they separate the effects of lactate carbon from the effects of pH in epigenetic reprogramming. This could be addressed in the Discussion.

https://doi.org/10.7554/eLife.50663.018

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Reviewer #1:

These authors studied the role of epigenetic reprogramming in cancer-associated fibroblasts (CAFs) from pancreatic ductal adenocarcinoma (PDAC). They find near-global DNA hypomethylation both in primary CAFs and in CAFs generated 'de novo' from mesenchymal stem cells cultivated in medium conditioned by PDAC cells. Hypomethylation was associated with widespread changes in gene expression, including the induction of numerous signaling genes. One such hypomethylated/overexpressed gene in CAFs was CXCR4, and siRNA silencing of this gene in CAFs was sufficient to suppress PDAC cell invasiveness. The authors then use 13C-base isotope tracing to show that lactate, a product secreted by PDAC cells, can be taken up by CAFs and used to supply the TCA cycle. 13C-lactate contributes to part of the CAF aKG pool, and this appeared to be associated with function of TET, an aKG-dependent enzyme, because inhibiting LDH suppressed TET activity. TET inactivation in MSCs also led to reduced production of CAFs and reduced xenograft growth in vivo. The authors conclude that lactate metabolism enables epigenetic reprogramming in CAFs, and that this in turn supports PDAC cell invasion and growth.

This area of work is potentially interesting because relatively little is known about metabolic cross-talk between cell types within the tumor microenvironment. The finding that TET is required for the MSC-to-CAF conversion is also interesting. However, there are several significant conceptual and methodological flaws with the paper in its current form.

1) In Figure 1, epigenetic marks are compared between CAFs and control stromal cells. This is problematic because the CAFs were generated from primary tumor samples and used at low passage, while the controls were purchased from ATCC and no information about passage number is provided. Given the known, potentially widespread epigenetic changes that occur in culture, the large changes reported in the figure could be due to culture-related artifacts. The type of culture medium, which is not described in detail for the cell lines used in the experiment, can also influence these marks. Overall, it is not clear that the authors are studying relevant changes in fibroblast biology.

The reviewer raises a good point. We used 4 distinct control fibroblasts and stellate cells that were low passage and were converted to DNA/RNA without growing them in vitro. We demonstrated loss of methylation in both primary CAFs as well as de novo generated CAFs. Now we have conducted whole genome hydroxymethylome and methylome analysis of CAFs (new Figure 4) and show that many important CAF related genes gain 5hmC marks. This is the first demonstration of genome wide hydoxymethylation in any tumor microenvironmental study. Additionally, we have also shown increased 5hmC by Mass Spec also. Taken together, all of these data point to reduced methylation and increased 5hmC in cancer associated fibroblasts.

2) Along these lines, with some 75% of the genome displaying altered cytosine methylation in CAFs, there is remarkably little overlap between promoters affected in primary and de novo-produced CAFs (Figure 2). This raises concern that the de novo CAFs are not particularly good models with which to study CAF epigenetics.

The analysis for the primary CAFs and de novo CAFs was done by two different platforms. There were different genomic loci that were represented on the HELP assay arrays that were used accounting for some of the discordance. Based on the reviewer’s question, we now conducted whole genome methylome and 5hmC analysis with oxidative bisulfite sequencing on MSCs and CAFs and demonstrate that virtually every CAF associated gene examined exhibits strong gains in 5hmC marks with concomitant loss of mC marks. These new data demonstrate that the loss of methylation in CAFs is now validated by many different assays.

3) In the experiments to manipulate CXCR4, only a single siRNA is used, silencing of mRNA levels does not appear very effective (only about 50% silencing in Figure 2D), no western blots are shown to document reduction in protein levels, and no attempts are made to reverse the effects (e.g. by expressing a non-silenced allele of CXCR4), making it hard to conclude that the siRNA's suppression of PDAC cell invasion are on target.

We have now used a second set of siRNAs with consistent results. We also show reduction at the protein level with immunoblotting (Figure 2—figure supplement 1). Our new 5hmC analysis also demonstrates acquisition of 5hmC not only at CXCR4 promoter, but also at the CXCR4 enhancer, strengthening the proposed hypothesis. Finally, we have now shown increased CXCR4 expression in the stromal cells in a set of 261 primary pancreatic cancer tissues (TMA) as shown in the new Figure 5.

4) In the metabolism experiments, the authors provide a high dose of 13C-lactate (10 mM) and culture the cells for 48 hours. Yet only 10% of the aKG pool is labeled. This observation provides strong evidence against the main conclusion of the paper: that lactate is an important source of aKG. Citrate is much more extensively labeled by lactate (40%), suggesting that most aKG arises from another carbon source, probably glutamine.

We thank the reviewers for making this observation, and we agree with them to a certain extent. However, the labeling of alpha-ketoglutarate (aKG) is difficult to take at face value as a representation of carbon contribution from any substrate. This is because aKG is a branch metabolite that connects the TCA cycle, anaplerotic glutamine flux and transamination reactions. Further, the conversion of Citrate to aKG, as well as Glutamate to aKG are reversible reactions. In the presence of glutamine and a source for pyruvate, equilibrium favors conversion of both Citrate and Glutamate to aKG. In order to further support our hypothesis, we repeated the 13-carbon labeled lactate experiment, with an additional condition of low glucose and low glutamine to highlight the contribution of lactate as a carbon substrate.

In these experiments as well as the experiments in our manuscript, the 13-carbon contribution from citrate is diluted by the 12-carbon from glutamate. This is also corroborated from the high 13-carbon labeling of alanine shown in Author Response Image 1, where the net production of 13-carbon labeled Alanine from pyruvate (derived from 13-carbon lactate) is accompanied by a net production of aKG from glutamate.

Author response image 1
Mass Isotopologue Distribution of Alanine from U-13C-Lactate tracing experiments indicate that lactate-derived carbon contributes significantly to de novo alanine synthesis.

The closest representation of carbon contribution from citrate to aKG, therefore, is represented by the total 13-carbon enrichment (mole percent enrichment, MPE) in citrate and aKG. The ration of aKG MPE to citrate MPE will indicate the net carbon contribution from 13-carbon lactate to citrate to aKG (Author response image 2).

Author response image 2

This data strongly supports our hypothesis that lactate is a significant carbon source for aKG, and this is especially evident under low glucose and low glutamine condition where lactate-derived citrate contributes to over 50% of the aKG carbon.

5) The authors should add lactate to the PDAC-conditioned medium in the LDH inhibitor experiment. This would provide evidence that the inhibitor's effect was on target. It would also be useful to perform this experiment genetically by silencing or mutating LDH in the CAFs – this should prevent lactate from being used as a carbon source in these cells.

We have now conducted LDH knockdown on MSC/CAFss and shown that this results in reduced expression of CAF conversion as evident from significantly reduced expression of FSP1, aSMA and VIM (Figure 3—figure supplement 2). Further experiments with exogenous lactate have been included in Figure 3—figure supplement 2.

6) Do the CAFs have any particular predilection to use lactate relative to the PDAC cells? This is an important question, because in the tumor microenvironment, these cells could presumably be competing for lactate.

This is an important point raised by the reviewer. In fact, we have recently shown that lactate secreted from cancer cells is preferably uptaken by CAFs (Nagrath et al., Cell Metab. 2016 Nov 8;24(5):685-700; Targeting Stromal Glutamine Synthetase in Tumors Disrupts Tumor Microenvironment-Regulated Cancer Cell Growth). In this work, we showed that CAFs used more lactate than glucose for acetyl-CoA synthesis.

Reviewer #2:

The present manuscript examines the links between the classical tumor cell metabolite lactate and CAF epigenetics in PDAC. The paper also zooms in on CXCR4 and TET2 as particular genes whose expression in CAFs is important for PDAC growth. The paper also proposes a mechanism by which lactate controls TET2 via KG. Overall, this paper examines a trendy idea (tumor cell secreted products impacting CAF epigenetics) and is concise and an appealing read, but I have serious concerns about the novelty, the quality, and especially the validity of the mechanism:

1) Figure 3 shows that labeled lactate labels TCA intermediates. This is something that was apparent to metabolism experts for ~ 50 years and shown beyond a shadow of a doubt by the Nat Chem paper from Patti lab.

We agree that it has been shown that lactate can lead to generation of TCA intermediates, but that biochemical observation was not the main message and novelty of our paper. We wanted to explore the molecular mechanisms of how CAFs are generated. Our paper is the first report to show epigenomic alterations during the process of MSCs to CAFs. Our paper is also the first report to point to the contribution of lactate in CAF generation. There have been recent reports on the contribution of lactate and other secreted agents on the “malignant cells”; but there have not been any reports demonstrating that lactate or any other secreted agent can lead reprogramming of MSCs to CAFs within the tumor micro-environment.

2) Labeling of KG is small (since most KG comes from glutamine in this [not fully physiologic] media). It is most unlikely that 10% of KG coming from lactate can have any affect on epigenetics. There is no evidence presented for altered KG concentration. It is the concentration that controls TET2 activity.

We have now shown that exogenous lactate leads to increased intracellular concentrations of aKG in CAFs (Figure 3—figure supplement 2) and that increased amounts of cell permeable aKG leads to increased TET activity (Figure 3—figure supplement 2). Furthermore, as discussed in Reviewer 1, point 4, our new metabolic experiments demonstrate the importance of lactate as a source of aKG (Figure 3—figure supplement 1).

3) The TET2 activity assay is in some version of lysates, which seemingly should not depend on cellular KG as the substrate (methods are not very complete, just referencing a kit). There is no clear logic connecting these biochemical TET2 measurements to KG or lactate. (Obviously lactate can impact cells many ways: pH, energy, pyruvate, NADH, etc.)

The TET activity assay was reliant on nuclear lysates generated from CAFs and MSCs by us. The nuclear lysate included the TET protein as well as other proteins, incl aKG that are present in the lysate. The activity is assessed by an ELISA that relies on ability of TET in converting mC to hmC. We have now added details on this assay to the methods section. We have used this assay in multiple recent papers that show TET activation by ascorbic acid (Shenoy, Verma et al., JCI 2019) and by phosphorylation (Jeong, Verma et al., Cancer Discovery, 2019, In Press).

4) P-values are reported for inappropriately small N (even N = 2).

The two assays that the reviewer is referencing to were de novo generation of CAFs from primary human MSC cultures. These were two independent biological experiments that rely on 21 day culture process in the presence of conditioned media. Due to the number of cells required (many millions) and the use of primary cells, we felt two independent biological replicates will be sufficient for the mass spec analysis of hmC. We have now conducted another 2 MSC to CAF conversion assays with biologically similar results and have shown them in supplementary figures.

5) It is hard to find concentrations of small molecule inhibitors employed, and there is no dose titration for lactate (used at 10 mM-- higher than reported values in PDAC tumors) or for small molecule tool compounds.

We used 10mM dose of Lactate based on published lactate concentrations in numerous studies in the stromal field and also from data generated from patient serum levels (Nature Cell Biology, (20), 597–609 (2018) Cancer-cell-secreted exosomal miR-105 promotes tumor growth through the MYC-dependent metabolic reprogramming of stromal cells; Walenta et al., Cancer Res. 2000 Feb 15;60(4):916-21; Semin Radiat Oncol. 2004 Jul; 14(3):267-74). Furthermore, a new study in eLife confirms high lactate levels in vivo in the same range (5mM to 10mM range) in pancreatic cancer and lung cancer models (Sullivan and Muir, eLife; https://elifesciences.org/articles/44235). The dose of LDH inhibitor, FX11 (10uM) was identical to the dose used in the original paper describing this specific inhibitor Chi Dang et al., Inhibition of lactate dehydrogenase A induces oxidative stress and inhibits tumor progression PNAS, 107(5), 2037–2042). We have now added these references in the text.

6) siRNA KD seem to be reported for single siRNA (evidence is that 5+ are needed for confirming target selectivity given frequent off-target effects) and without rescue (which is of great importance for proving on-target mechanism); there is no use of more advanced and reliable genetics.

We now show data with second set of siRNAs and also show reduction at the level of protein expression (Figure 2—figure supplement 1). Also, we have now shown 5hmC acquisition and CXCR4 overexpression in a cohort of 250+ primary PDAC sample, in single cell RNA-seq studies as well as in the KPC mouse model (new Figure 5).

7) There are no experiments actually testing paracrine lactate-mediated epigenetic reprogramming in a physiological context.

In summary, the paper is nicely constructed but limited in conceptual novelty and does not come close to proving the proposed mechanism from lactate-->KG-->Tet-->hypomethylation-->CXCR4 expression.

As mentioned in response to point no 5, we based the lactate concentrations on multiple published studies. In fact, a new study in eLife confirms high lactate levels in vivo in the same range (5mM to 10mM range) in pancreatic cancer and lung cancer models (Sullivan and Muir, eLife; https://elifesciences.org/articles/44235). We have now shown that TET activity increases in a dose dependent fashion when MSCs are exposed to exogenous lactate (Figure 3—figure supplement 2). We also show the role of TET enzymes in CAF formation in a mouse model in vivo (Figure 5).

Reviewer #3:

This is an excellent study highlighting the role of metabolism reprograming on epigenetics of stromal cells to control tumor growth. I have a few experiments to bolster their conclusions.

Experiments:

1) The authors suggest that lactate is a fuel to support TCA cycle metabolite generation. I assume lactate becomes pyruvate that enters the mitochondria where it is converted into acetyl-CoA via PDH. The authors should prevent pyruvate entry into the mitochondria by using mitochondrial pyruvate carrier inhibitor UK5099 and examine methylation status.

Based on the reviewers suggestion we have performed experiments using pyruvate inhibitor UK5099 and show that this inhibition reduces the TET2 activation as well as 5hmC increase that is seen by exogenous lactate in CAF cells (Figure 3—figure supplement 2).

2) Does cell permeable aKG recapitulate the effects of lactate (2 weeks) on methylation in CAFs?

It has also been shown previously that cell permeable aKG does lead to TET activation and increased 5hmC (Carey et al., 2015). Based on the reviewer’s suggestion, we used a cell permeable aKG and show that TET2 activity does increase with this agent in CAF cells (Figure 3—figure supplement 2).

3) It is possible that the presence of lactate diminishes glycolytic flux, which could diminish one-carbon metabolism (SAM levels). They should measure SAM levels as well as provide ratios of succinate/aKG, 2HG/aKG and fumarate/aKG, which have been shown to inhibit TETs, in CAFs +/-lactate.

The reviewer brings out an excellent observation and we thank them for bringing this to our attention. With regards to lactate reducing glycolytic flux and subsequently affecting SAM cycle, unfortunately, direct measurement of S-adenosyl methionine is difficult with GC-MS techniques used for measuring polar metabolites. However, we looked at serine and glycine levels in the presence of lactate, since glycolytic carbon to SAM cycle is routed through serine and glycine. As the reviewer anticipated, we do find an indication that glycolytic flux to SAM cycle might be decreasing by observing the Serine:Pyruvate abundance ratio (Figure 3—figure supplement 1).

As for the metabolite ratio being indicative of TET activity, we calculated 2-HG:aKG and Fumarate:aKG ratio in media with and without Lactate. The 2-HG signals were low across all the samples, which was expected as 2-HG accumulation is prevalent in the presence of mutated IDH and there is no evidence IDH mutations in these cells. The small amounts of 2-HG detected are presumably from the promiscuous IDH activity. Despite this, the 2-HG:aKG ratio in our observations tend to decrease when lactate is present in the media, indicating higher TET activity.

More interestingly, Fumarate:aKG ratio drops markedly in the presence of lactate, indicating decrease in TET inhibition as observed in the studies by Xiao et al. and Laukka et al. Unfortunately, total succinate levels could not be estimated correctly due to succinate and proline co-eluting in our GC-MS analysis.

Xiao M, Yang H, Xu W, et al. Inhibition of α-KG-dependent histone and DNA demethylases by fumarate and succinate that are accumulated in mutations of FH and SDH tumor suppressors. Genes Dev. 2012;26(12):1326-38.

Laukka, T., Mariani, C.J., Ihantola, T., Cao, J.Z., Hokkanen, J., Kaelin, W.G., Godley, L.A., and Koivunen, P. (2016). Fumarate and Succinate Regulate Expression of Hypoxia-inducible Genes via TET Enzymes. J. Biol. Chem. 291, 4256–4265.

All of these results are now shown in Figure 3—figure supplement 1 of the revised manuscript.

4) Can they recapitulate the KPC + TET2 KO CAF tumor experiments with KPC + WT CAFs treated with lactate for 2 weeks (Figure 4G)?

We appreciate the reviewer’s point. Our aim was to show that MSCs from TET2 KO mice will not be able to form functional CAFs and will not be able to support tumor growth as well as the WT CAFs. Based on the reviewer’s advice we did treat WT CAFs with exogenous lactate. Unfortunately, these primary mouse derived cells did not grow very well in vitro and we did not achieve sufficient numbers to perform the in vivo xenograft studies.

[Editors' note: the author responses to the re-review follow.]

Essential revisions:

1) There needs to be more clarity about the assays used to measure TET activity in lysates. The authors are proposing that aKG abundance from the lysate influences the activity reported by this commercial assay. But does the assay buffer contain aKG? If so, it seems unlikely that the relatively small amount of aKG in the lysates could substantially affect TET activity. Please clarify: (i) The source and abundance of aKG in the assay, which is not clear from the methods or the vendor website; and (ii) based on the answer, a clear delineation in a few sentences of reasonable possibilities of how components in the lysate (including aKG) might influence the results of the assay.

We were able to obtain information from the vendor (Epigentek) that the TET activity assay buffer does contain cofactors required for TET enzymatic activity: Ascorbic Acid (1mM), aKG (1mM) and FeNH4SO4 (0.1mM). The assay does not contain TET though and thus relies on TET obtained from the nuclear cell lysates. The assay is also developed very quickly (5 minutes). The assay relies on the conversion of 5mC to 5hmC by TET enzymatic activity. Notwithstanding that the assay contains aKG; we were still able to obtain a significant difference in TET activity between the CAFs and MSCs, suggesting that CAFs potentially contained activated TET that was able to convert mC to hmC in a short period of time. We have now added text about the constituents of the assay and the inferred conclusions in the revised text.

2) In the original submission, two reviewers questioned the importance of lactate for formation of aKG when isotope tracing with 13C-lactate resulted only in very low labeling of the aKG pool. The explanation for this result seems to be that lactate-derived synthesis of alanine results in production of unlabeled aKG; therefore observation of labeled aKG underestimates lactate's contribution. If this is the argument, it should be spelled out clearly in the manuscript itself, because the impression is still that lactate makes only a small contribution to aKG. The authors also need to explain why the fractional enrichment in alanine is so different between Figure 3A and Figure 3—figure supplement 1A. The latter was used to make the point about alanine labeling in the rebuttal, but the enrichment in both high and low glucose conditions in this experiment is much higher than in the main figure. Finally, it would help the reader if the "accessory" formation of aKG via alanine synthesis (i.e. that pyruvate's conversion to alanine results in glutamate's conversion to aKG) was clearly illustrated in the model in Figure 3F.

The reviewer brings up an important point. The fractional enrichment of alanine in the main figure and supplementary figures is disparate due to the different media formulations used. The data in the main figure corresponds to MSCs cultured in aplhaMEM with 13C-lactate, whereas the supplementary figure shows results for MSCs cultured in RPMI with 13C-lactate. Since α-MEM contains unlabeled alanine, the 13C-enrichment of intracellular alanine is diluted and appears less than the corresponding enrichment of alanine in RPMI media, which does not contain alanine. The MSCs were cultured in both α-MEM and RPMI to highlight the lactate-derived synthesis of alanine via GPT that was being masked in α-MEM cultures. In addition, we have amended the schematic to show the accessory formation of aKG via GPT in revised Figure 3A.

3) Evidence for paracrine lactate exchanges in vivo is still circumstantial. Some of the same phenomena are observed in culture and in vivo, and the evidence for TET2's involvement in vivo is strong. But there is still no compelling evidence that lactate released by cancer cells has anything to do with TET2 activity in CAFs in the tumor microenvironment. Adding direct evidence along these lines would strengthen the paper. Failing that, the authors need to be more circumspect about the role of this mechanism in tumors.

Based on the reviewers’ suggestion, we have changed the language in Abstract, Introduction and Discussion in the paper as follows, to alleviate the reviewer’s concerns:

Abstract: We have added “associated” instead of “results” and “seen in” instead of “regulates”

Previous line: Thus, in PDAC, a tumor-mediated lactate flux results in widespread epigenomic reprogramming that regulates CAF formation

Revised line: Thus, in PDAC, a tumor-mediated lactate flux is associated with widespread epigenomic reprogramming that is seen during CAF formation.

In the Introduction: “Importantly, from a mechanistic standpoint, we determine that paracrine lactate secreted by PDAC cells can be incorporated in stromal cells and lead to increased alpha-keto glutarate (aKG). This is associated with activation of the TET demethylase, thus potentially leading to epigenetic reprogramming seen during CAF formation.

We have made similar changes in the Discussion also.

4) The authors should address how they separate the effects of lactate carbon from the effects of pH in epigenetic reprogramming. This could be addressed in the Discussion.

This is an excellent point raised by the reviewer. In fact, we did find one manuscript that showed that low acidic ph leads to increases in 2HG more than aKG (Nadtochiy et al., 2016). 2HG increase generally leads to decreased Tet activity. Thus in our model, a low ph (due to increase lactic acid) should not account for the high Tet activity and consequent decreased 5mC that we see. We have added this in the Discussion also.

https://doi.org/10.7554/eLife.50663.019

Article and author information

Author details

  1. Tushar D Bhagat

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Conceptualization, Data curation, Methodology, Project administration
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4527-5505
  2. Dagny Von Ahrens

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  3. Meelad Dawlaty

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Resources, Data curation
    Competing interests
    No competing interests declared
  4. Yiyu Zou

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  5. Joelle Baddour

    Department of Biomedical Engineering, University of Michigan, Ann Arbor, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  6. Abhinav Achreja

    Department of Biomedical Engineering, University of Michigan, Ann Arbor, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  7. Hongyun Zhao

    Department of Biomedical Engineering, University of Michigan, Ann Arbor, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  8. Lifeng Yang

    Department of Biomedical Engineering, University of Michigan, Ann Arbor, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  9. Brijesh Patel

    Rutgers University, New Brunswick, United States
    Contribution
    Data curation, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  10. Changsoo Kwak

    1. Department of Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    2. Department of Translational Molecular Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    Contribution
    Data curation, Investigation
    Competing interests
    No competing interests declared
  11. Gaurav S Choudhary

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  12. Shanisha Gordon-Mitchell

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Resources, Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  13. Srinivas Aluri

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Funding acquisition, Investigation
    Competing interests
    No competing interests declared
  14. Sanchari Bhattacharyya

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation
    Competing interests
    No competing interests declared
  15. Srabani Sahu

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation
    Competing interests
    No competing interests declared
  16. Prafulla Bhagat

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation; Investigation; Methodology
    Competing interests
    No competing interests declared
  17. Yiting Yu

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  18. Matthias Bartenstein

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0908-770X
  19. Orsi Giricz

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Investigation
    Competing interests
    No competing interests declared
  20. Masako Suzuki

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Resources, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  21. Davendra Sohal

    Department of Medicine, Cleveland Clinic, Cleveland, United States
    Contribution
    Formal analysis, Methodology
    Competing interests
    No competing interests declared
  22. Sonal Gupta

    1. Department of Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    2. Department of Translational Molecular Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    Contribution
    Resources, Investigation, Methodology
    Competing interests
    No competing interests declared
  23. Paola A Guerrero

    1. Department of Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    2. Department of Translational Molecular Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    Contribution
    Resources, Data curation, Investigation
    Competing interests
    No competing interests declared
  24. Surinder Batra

    University of Nebraska Medical Center, Omaha, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  25. Michael Goggins

    Johns Hopkins, Baltimore, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  26. Ulrich Steidl

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Formal analysis, Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  27. John Greally

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6069-7960
  28. Beamon Agarwal

    GenomeRxUs LLC, Secane, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  29. Kith Pradhan

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  30. Debabrata Banerjee

    Rutgers University, New Brunswick, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  31. Deepak Nagrath

    Biointerfaces Institute, University of Michigan, Ann Arbor, United States
    Contribution
    Data curation, Formal analysis, Project administration, Writing—review and editing
    For correspondence
    dnagrath@umich.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8999-2282
  32. Anirban Maitra

    1. Department of Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    2. Department of Translational Molecular Pathology, Sheikh Ahmed Pancreatic Cancer Research Center, UT MD Anderson Cancer Center, Houston, United States
    Contribution
    Conceptualization, Formal analysis, Project administration, Writing—review and editing
    For correspondence
    AMaitra@mdanderson.org
    Competing interests
    No competing interests declared
  33. Amit Verma

    Albert Einstein College of Medicine, Montefiore Medical Center, New York, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    amit.verma@einstein.yu.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7592-7693

Funding

Albert Einstein College of Medicine (Einstein Training Program in Stem Cell Research C34874GG)

  • Tushar D Bhagat

National Cancer Institute (T32 CA200561)

  • Dagny Von Ahrens

National Cancer Institute (R01CA227622)

  • Deepak Nagrath

National Cancer Institute (R01CA222251)

  • Deepak Nagrath

National Cancer Institute (R01CA204969)

  • Deepak Nagrath

National Cancer Institute (R01CA218004)

  • Anirban Maitra

National Cancer Institute (P01CA117969)

  • Anirban Maitra

National Cancer Institute (R01CA220236)

  • Anirban Maitra

National Cancer Institute (U24CA224020)

  • Anirban Maitra

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

TB was supported by The Einstein Training Program in Stem Cell Research from the Empire State Stem Cell Fund through New York State Department of Health Contract C34874GG. DVA was supported by a CA200561 T32 training grant. DN is supported by grants from NCI R01CA227622, R01CA222251, and R01CA204969

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC no. 20181208) protocols of the Albert Einstein College of Medicine.

Senior Editor

  1. Anna Akhmanova, Utrecht University, Netherlands

Reviewing Editor

  1. Ralph DeBerardinis, UT Southwestern Medical Center, United States

Version history

  1. Received: August 21, 2019
  2. Accepted: October 27, 2019
  3. Accepted Manuscript published: October 30, 2019 (version 1)
  4. Accepted Manuscript updated: November 1, 2019 (version 2)
  5. Version of Record published: November 22, 2019 (version 3)

Copyright

© 2019, Bhagat et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 4,858
    Page views
  • 883
    Downloads
  • 95
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Tushar D Bhagat
  2. Dagny Von Ahrens
  3. Meelad Dawlaty
  4. Yiyu Zou
  5. Joelle Baddour
  6. Abhinav Achreja
  7. Hongyun Zhao
  8. Lifeng Yang
  9. Brijesh Patel
  10. Changsoo Kwak
  11. Gaurav S Choudhary
  12. Shanisha Gordon-Mitchell
  13. Srinivas Aluri
  14. Sanchari Bhattacharyya
  15. Srabani Sahu
  16. Prafulla Bhagat
  17. Yiting Yu
  18. Matthias Bartenstein
  19. Orsi Giricz
  20. Masako Suzuki
  21. Davendra Sohal
  22. Sonal Gupta
  23. Paola A Guerrero
  24. Surinder Batra
  25. Michael Goggins
  26. Ulrich Steidl
  27. John Greally
  28. Beamon Agarwal
  29. Kith Pradhan
  30. Debabrata Banerjee
  31. Deepak Nagrath
  32. Anirban Maitra
  33. Amit Verma
(2019)
Lactate-mediated epigenetic reprogramming regulates formation of human pancreatic cancer-associated fibroblasts
eLife 8:e50663.
https://doi.org/10.7554/eLife.50663

Further reading

    1. Cancer Biology
    Gehad Youssef, Luke Gammon ... Adrian Biddle
    Research Article

    Cancer stem cells (CSCs) undergo epithelial-mesenchymal transition (EMT) to drive metastatic dissemination in experimental cancer models. However, tumour cells undergoing EMT have not been observed disseminating into the tissue surrounding human tumour specimens, leaving the relevance to human cancer uncertain. We have previously identified both EpCAM and CD24 as CSC markers that, alongside the mesenchymal marker Vimentin, identify EMT CSCs in human oral cancer cell lines. This afforded the opportunity to investigate whether the combination of these three markers can identify disseminating EMT CSCs in actual human tumours. Examining disseminating tumour cells in over 12,000 imaging fields from 74 human oral tumours, we see a significant enrichment of EpCAM, CD24 and Vimentin co-stained cells disseminating beyond the tumour body in metastatic specimens. Through training an artificial neural network, these predict metastasis with high accuracy (cross-validated accuracy of 87-89%). In this study, we have observed single disseminating EMT CSCs in human oral cancer specimens, and these are highly predictive of metastatic disease.

    1. Cancer Biology
    2. Medicine
    Dingyu Rao, Hua Lu ... Defa Huang
    Research Article

    Esophageal cancer (EC) is a fatal digestive disease with a poor prognosis and frequent lymphatic metastases. Nevertheless, reliable biomarkers for EC diagnosis are currently unavailable. Accordingly, we have performed a comparative proteomics analysis on cancer and paracancer tissue-derived exosomes from eight pairs of EC patients using label-free quantification proteomics profiling and have analyzed the differentially expressed proteins through bioinformatics. Furthermore, nano-flow cytometry (NanoFCM) was used to validate the candidate proteins from plasma-derived exosomes in 122 EC patients. Of the 803 differentially expressed proteins discovered in cancer and paracancer tissue-derived exosomes, 686 were up-regulated and 117 were down-regulated. Intercellular adhesion molecule-1 (CD54) was identified as an up-regulated candidate for further investigation, and its high expression in cancer tissues of EC patients was validated using immunohistochemistry, real-time quantitative PCR (RT-qPCR), and western blot analyses. In addition, plasma-derived exosome NanoFCM data from 122 EC patients concurred with our proteomic analysis. The receiver operating characteristic (ROC) analysis demonstrated that the AUC, sensitivity, and specificity values for CD54 were 0.702, 66.13%, and 71.31%, respectively, for EC diagnosis. Small interference (si)RNA was employed to silence the CD54 gene in EC cells. A series of assays, including cell counting kit-8, adhesion, wound healing, and Matrigel invasion, were performed to investigate EC viability, adhesive, migratory, and invasive abilities, respectively. The results showed that CD54 promoted EC proliferation, migration, and invasion. Collectively, tissue-derived exosomal proteomics strongly demonstrates that CD54 is a promising biomarker for EC diagnosis and a key molecule for EC development.