1. Cancer Biology
  2. Cell Biology
Download icon

Human pancreatic cancer cell exosomes, but not human normal cell exosomes, act as an initiator in cell transformation

  1. Karoliina Stefanius
  2. Kelly Servage
  3. Marcela de Souza Santos
  4. Hillery Fields Gray
  5. Jason E Toombs
  6. Suneeta Chimalapati
  7. Min S Kim
  8. Venkat S Malladi
  9. Rolf Brekken
  10. Kim Orth  Is a corresponding author
  1. University of Texas Southwestern Medical Center, United States
  2. Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, United States
Research Communication
  • Cited 2
  • Views 2,405
  • Annotations
Cite this article as: eLife 2019;8:e40226 doi: 10.7554/eLife.40226

Abstract

Cancer evolves through a multistep process that occurs by the temporal accumulation of genetic mutations. Tumor-derived exosomes are emerging contributors to tumorigenesis. To understand how exosomes might contribute to cell transformation, we utilized the classic two-step NIH/3T3 cell transformation assay and observed that exosomes isolated from pancreatic cancer cells, but not normal human cells, can initiate malignant cell transformation and these transformed cells formed tumors in vivo. However, cancer cell exosomes are unable to transform cells alone or to act as a promoter of cell transformation. Utilizing proteomics and exome sequencing, we discovered cancer cell exosomes act as an initiator by inducing random mutations in recipient cells. Cells from the pool of randomly mutated cells are driven to transformation by a classic promoter resulting in foci, each of which encode a unique genetic profile. Our studies describe a novel molecular understanding of how cancer cell exosomes contribute to cell transformation.

Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that major issues remain unresolved (see decision letter).

https://doi.org/10.7554/eLife.40226.001

Introduction

Within the tumor microenvironment, a dynamic molecular communication between tumor and surrounding stromal cells is a well-recognized feature of cancer progression (Salvatore et al., 2017; Kahlert and Kalluri, 2013). The conversation at the primary tumor site, as well as at distant locations, is mediated through many secreted factors including exosomes; small (30–150 nm) secreted extracellular vesicles shed by normal and malignant cells (Kalluri, 2016; Costa-Silva et al., 2015; Stahl and Raposo, 2018; Colombo et al., 2014; Willms et al., 2016). Exosomes and their mechanism(s) of biogenesis and function has emerged as a promising, yet controversial, field of research. Based on multi-omic studies, exosomes are known to carry heterogeneous cargo composed of proteins, metabolites, genetic material (DNA and microRNAs), and lipids (Willms et al., 2016; Demory Beckler et al., 2013; Kowal et al., 2016; Kalluri and LeBleu, 2016; Melo et al., 2014; Smith et al., 2015). They are selectively packaged and transferred into recipient cells acting as vehicles for intercellular communication in normal physiological and pathological conditions (Hessvik and Llorente, 2018). A growing body of evidence shows exosomes are crucial in shaping the local tumor microenvironment to promote cancer progression by advancing tumor metastasis (Costa-Silva et al., 2015; Zhang et al., 2015). Although there is a major emphasis on describing the function of exosomes in metastasis and interactions in the local tumor microenvironment, less effort has been invested in analyzing their specific contribution to transforming a normal cell into a malignant cell. It has been shown that exosomes can contribute to the transformation of nontumorigenic cells to form tumors (Melo et al., 2014; Dai et al., 2018; Abdouh et al., 2017; Hamam et al., 2016; Antonyak et al., 2011). However, a common feature among these studies is the use of cells that were either pre-exposed to transforming agents or treated with cancer patient sera or medium from cultured cancer cells, both of which contain cancer cell exosomes. Under these conditions, it is conceivable that other components, in addition to exosomes, are transforming cells.

Pancreatic cancer is a lethal metastatic disease that lacks efficient curative treatment. Even though there is continuing progress toward understanding the biology of pancreatic cancer, it remains one of the leading causes of cancer-related deaths in the world (Vincent et al., 2011). This is mainly due to the lack of effective treatments against its highly metastatic behavior and therefore, understanding the key mechanisms underlying its progression is needed. Mutations in the four pancreatic driver genes, KRAS, CDKN2A, TP53, and SMAD4, occur frequently in pancreatic ductal adenocarcinomas (PDAC) and are well described (Giovannetti et al., 2017). An activating mutation in the KRAS gene is present in 90% of cases (Giovannetti et al., 2017). Additionally, genetic heterogeneity and polyclonality have also been shown to be present in PDAC (Giovannetti et al., 2017). Together with the indications that cancer-cell-derived exosomes are emerging contributors to tumor promotion, we wanted to evaluate whether exosomes secreted by pancreatic cancer cells participate in a distinct role in the process of cell transformation.

Malignant transformation of a normal cell occurs in a stepwise fashion. Point mutations in the genome can result in the reprogramming of a normal cell to a less differentiated state that is receptive to additional genetic alterations resulting in uncontrolled growth and ultimately cancer. The classic two-stage in vitro cell transformation assay (CTA) is a tiered system for transformation that was created for screening potential carcinogenic factors (Berwald and SACHS, 1963; Kakunaga, 1973; Sakai and Sato, 1989). In this system, cells are first treated with a suspected carcinogen, called an initiator, such as the genotoxic carcinogen 3-MCA (3-methylcholanthrene). 3-MCA introduces random genetic changes in a pool of normal cells. Subsequently, these initiated cells are exposed to a promoter, such as TPA (12-O-tetradecanoylphorbol 13-acetate), which enhances proliferation in the initiated cells selectively, thus driving malignant transformation of the cells. The resulting transformed cells are observed as foci on a cell culture plate (Sakai and Sato, 1989; Sasaki et al., 2012). This reductionist approach provides sensitivity in detecting a wider range of initiating agents that may not show obvious transforming activity without a promoter (Sakai and Sato, 1989). Using this assay as a model system for malignant cell transformation, we assessed whether or not cancer-cell-derived exosomes could affect and/or potentially drive the transformation of a normal cell.

The results presented herein provide a detailed analysis of a previously unidentified molecular function of cancer cell exosomes for malignant cell transformation. We observe that exosomes derived from pancreatic cancer cells can act as an initiator but not as a promoter in the two-stage CTA leading to malignant cell transformation. By contrast, exosomes derived from normal pancreatic cells have no effect on the cell transformation process. Specifically, using this two-stage CTA, we observe over a three-day initiator step that a single treatment of cancer cell exosomes acts in the same manner as a single treatment of the chemical initiator 3-MCA. As initiators of cell transformation, they incorporate random molecular changes into DNA. These random mutations then set the stage for a promoter to induce transformation of cells to form foci. In addition, we show that the cancer cell exosome-initiated transformed cells form aggressive tumors when injected into mice. In-depth analysis of the transformed cells using a combination of proteomics and exome sequencing reveals distinct differences between healthy cells and transformed cells with key insights into how cancer cell exosomes may be working on recipient cells to contribute to transformation. Overall, this study uncovers a specific function of pancreatic cancer cell exosomes in the malignant cell transformation process.

Results

Exosome isolation, validation, and characterization

Exosomes were isolated using a combined ultrafiltration-ultracentrifugation protocol (described in detail in Materials and methods) from four pancreatic cancer cell lines, Capan-2, MIA PaCa-2, Panc-1, and BxPC-3, and two human normal cell lines, human pancreatic ductal epithelial cells (HPDE) and human primary dermal fibroblasts (Adamczyk et al., 2011). Of note, three of the cancer cell lines, Capan-2, MIA PaCa-2, and Panc-1, have oncogenic mutations in the KRAS gene, whereas BxPC-3 has wild-type KRAS (Deer et al., 2010). To confirm rigor and reproducibility, isolated exosomes from each cell type were characterized for the presence of common exosome-associated proteins using mass spectrometry and immunoblot analysis (Figure 1, Figure 1—figure supplement 1, Figure 1—figure supplement 2). Additionally, electron microscopy (TEM) and nanoparticle tracking analysis (NTA) were employed to analyze the morphology and size range of isolated exosomes (Figure 1, Figure 1—figure supplement 1) (Willms et al., 2016; Kowal et al., 2016; Lötvall et al., 2014; Witwer et al., 2017; Théry et al., 2018). An example of this characterization of exosomes is shown for Capan-2 exosomes in Figure 1. Immunoblot analysis was used to confirm the presence of expected exosomal marker proteins (CD63, Alix, and TSG101) as well as the absence of proteins not commonly found in exosomes (Calnexin, α-actinin, and HSP90) (Figure 1A,B). In addition to western blot analysis, mass spectrometry analysis of isolated exosomes confirms the presence of the top twenty most commonly found proteins in exosomes according to the ExoCarta database (Figure 1—figure supplement 2) (Keerthikumar et al., 2016). TEM images of exosomes isolated from Capan-2 cells show the expected round or cup-shaped morphology (Willms et al., 2016) and NTA shows the size distribution of exosomes centered on 91 nm with a mean size of 250.3 nm (Figure 1C,D).

Figure 1 with 2 supplements see all
Exosome isolation, validation, and characterization.

(A) Western blot analysis of common exosomal marker proteins CD63, Alix, and TSG10 found in exosomes isolated from Capan-2 cells. (B) Western blot analysis of proteins HSP90, Calnexin, and α-actinin, expected to be underrepresented in exosomes. Equivalent amounts of proteins from P2 (ER and mitochondria), S2 (cytoplasm), M (media), and Ex (exosome) fractions derived from the Capan-2 exosome isolation process were loaded into gel for the analysis. (C) Nanoparticle Tracking Analysis of ‘crude’ Capan-2 cell exosomes. Data represent average size per concentration (black line) ± standard error of the mean (red bars) of three measurements from one exosome preparation. Exosome size is centered on 91 nm with a mean size of 250.3 nm. Finite Track Length Analysis (FTLA) was used for size determination. (D) Representative TEM images of exosomes isolated from Capan-2 cells shown at three different scales confirm expected cup-shaped morphology of vesicles.

https://doi.org/10.7554/eLife.40226.002

Pancreatic cancer cell exosomes function as an initiator in malignant cell transformation

To analyze if exosomes contribute to malignant cell transformation, we utilized the two-stage CTA as a model system for transformation. The CTA was performed with NIH/3T3 cells using an established chemical initiator and promoter, MCA and TPA, respectively (Sakai and Sato, 1989; Alvarez et al., 2014). As shown in Figure 2A, the complete assay is 42 days long and involves the treatment of NIH/3T3 cells with an initiator (3 days) followed by a promoter (2 weeks) before recovery (3 weeks). Successful cell transformation results in the formation of foci that are identified by defined criteria described in Materials and methods (Figure 2B). Cells that are untreated, treated with only an initiator (MCA), or only a promoter (TPA) show the formation of low levels of background foci (one foci/well on average) (Figure 2C, Figure 2—source data 1) (Sasaki et al., 2012). By contrast, cells treated with both MCA as the initiator and TPA as the promoter resulted in formation of 3–4 foci/well on average (Figure 2C, Figure 2—source data 1). Consistent with previous findings, treatment with both an initiator and promoter was required to observe increased cell transformation as evidenced by formation of an increased number of foci above background levels (Sakai and Sato, 1989).

Figure 2 with 1 supplement see all
Pancreatic cancer cell exosomes function as an initiator in malignant cell transformation.

(A) Two-stage cell transformation assay shown. NIH/3T3 cells were treated with a tumor initiator for 3 days (Days 3–6) and the tumor promoter for 2 weeks (Days 8–21). After 42 days, cells are fixed with methanol and stained with Crystal Violet for malignant foci counting. (B) Representative images of stained cells showing foci formation (arrows) from untreated cells and cells treated with MCA/DMSO, DMSO/TPA, MCA/TPA, or Capan-2 exosomes (ExC)/TPA (initiator/promoter). (C) Quantification of foci formed at the end of cell transformation assays. The average foci/well were determined via double-blind counting as described in Materials and methods. The red dashed line represents the established level of background foci present in untreated cells. Initiator/promoter treatments resulting in increased foci formation above background include MCA/TPA (p=0.008) and all cancer cell-derived exosomes: ExC/TPA (p=0.0002), ExM/TPA (p<0.0001), ExP/TPA (p=0.007), and ExBx/TPA (p=0.0003). Bars shown in gray represent controls that did not result in foci formation above background. Bar shown in pink shows results from normal cell (HPDE) exosome/TPA treatment (p=0.0004). (D) Quantification of foci formed after use of a different promoter, CdCl2. CdCl2 acts as a promoter leading to increased foci formation above background when used with the initiators MCA (p<0.0001) or Capan-2 exosomes (ExC) (p<0.0001). Asterisks indicate significant differences from either control treatment or MCA/TPA treatment as determined by unpaired, two-tailed t-test with Welch’s correction (*p<0.05; **p<0.01; ***p<0.001; ****p<0.0001). Red (+)=initiator used; purple (+)=promoter used.

https://doi.org/10.7554/eLife.40226.007

To establish what role, if any, cancer-cell exosomes have on cell transformation, we assessed whether exosomes isolated from three pancreatic cancer cell lines, Capan-2, MIA PaCa-2, and Panc-1, which are known to carry oncogenic mutations in the KRAS gene, could act as an initiator and/or promoter in the CTA. NIH/3T3 cells were first treated with isolated exosomes from each of the three pancreatic cancer cell lines for the duration of the initiation and promotion steps (3-week treatment). This resulted in the formation of only background transformation activity, similar to what is observed with the untreated control (Figure 2C, Figure 2—source data 1). Next, NIH/3T3 cells were first treated with the initiator MCA and then subsequently treated with isolated cancer cell exosomes for the 2-week promotion period. For each of these three cancer cell exosome assays, again only background levels of foci were observed (Figure 2C, Figure 2—source data 1). However, when the cancer cell exosomes were tested as an initiator in combination with the promoter TPA, cell transformation was observed at similar levels as the chemical MCA/TPA treatment (3–4 foci/well) (Figure 2C, Figure 2—source data 1). Therefore, exosomes derived from three different pancreatic cancer cell lines (Capan-2, MIA PaCa-2 or Panc-1) can each function as an initiator in the CTA resulting in transformation of NIH/3T3 cells.

To assess whether this initiator activity is a general characteristic of exosomes from all cell types or a trait unique to exosomes derived from KRAS mutated pancreatic cancer cell lines, we repeated experiments with exosomes from three additional cell types: BxPC-3 cells, a pancreatic cancer cell line with WT KRAS, HPDE cells, a normal human cell line, and primary dermal fibroblasts. We observed that cancer cell exosomes isolated from WT KRAS BxPC-3 cells can act as an initiator of cell transformation, resulting in foci formation similar to the numbers observed with MCA/TPA treatment (Figure 2C, Figure 2—source data 1). However, normal cell exosomes isolated from HPDE cells or exosomes from primary fibroblasts were unable to induce cell transformation when used as an initiator in the CTA (Figure 2C, Figure 2—source data 1, Figure 2—figure supplement 1A, Figure 2—figure supplement 1—source data 1). Collectively, the results show that pancreatic cancer cell exosomes can act as an initiator in malignant cell transformation of NIH/3T3 cells, while exosomes isolated from normal pancreatic cells or primary fibroblasts cannot.

Gradient purified exosomes contain initiator activity

Exosomes were isolated using an ultrafiltration-ultracentrifugation method and validated using a number of criteria (Figure 1). Our results demonstrate a specific function for cancer cell exosomes, but we used a protocol that is known to result in a preparation containing both exosomes and aggregated protein/nucleic acid contaminants. For this reason, we performed an additional purification step by floating exosomes onto a sucrose density gradient to obtain cleaner exosome preparation separated from contaminants (Chiou and Ansel, 2016). Characterization of these purified exosomes (Fraction 3) by NTA showed a size range centered on 67 nm with a mean size of 83.5 nm. Immunoblot analysis confirmed the presence of expected exosomal marker proteins CD63 and Alix. In addition, mass spectrometry analysis confirmed the presence of the top twenty most commonly found proteins in exosomes according to the ExoCarta database (Figure 1—figure supplement 1, Figure 1—figure supplement 2). The purified exosomes were then tested as an initiator with the promoter TPA in the transformation assay and results show that the population of ‘pure’ exosomes retain the ability to act as an initiator of cell transformation (Figure 2—figure supplement 1B, Figure 2—figure supplement 1—source data 2).

Cancer cell exosome initiator activity is detected at low concentrations, requiring intact exosomes

Dose-response studies were performed using protein concentration as a normalization strategy to evaluate the amount of cancer cell exosomes needed to initiate cell transformation. As a standard concentration in each transformation assay, we used 80 ng/mL of proteins, corresponding to 7.0 × 107 particles/mL. Exosome protein concentrations ranging from 0.08 ng/mL to 2400 ng/mL were tested and we observed equal cell transformation for all concentrations with the exception of the two lowest, 0.08 ng/mL and 0.8 ng/mL. This indicates that initiator activity of cancer cell exosomes requires one dose of exosomes over a 3-day period with a protein concentration of at least 8 ng/mL (Figure 2—figure supplement 1C, Figure 2—figure supplement 1—source data 3). Furthermore, when cancer cell exosomes are boiled for 10 min at 100°C just prior to use as an initiator, the level of transformed foci decreases to background levels (Figure 2—figure supplement 1A, Figure 2—figure supplement 1—source data 1).

Cancer-cell-derived exosomes function as a classic initiator

One common characteristic of initiators in the CTA is the capability of working with multiple promoters to induce cell transformation (Fang et al., 2001; Sakai, 2007). To test for this with the pancreatic cancer cell exosomes, we replaced TPA with another common promoter, cadmium chloride (CdCl2) (Fang et al., 2001; Keshava et al., 2000; Umeda et al., 1989). We observed that NIH/3T3 cells treated with either MCA or Capan-2 exosomes as the initiator followed by treatment with CdCl2 as the promoter resulted in formation of foci similar to that observed with TPA as the promoter (Figure 2D, Figure 2—source data 2). Treatment of cells with CdCl2 alone resulted in background levels of foci, reiterating the fact that cell transformation is dependent on both initiation and promotion. These results indicate that exosomes isolated from pancreatic cancer cells act as general initiators in the transformation assay and are not dependent on a specific promoter.

In vivo studies confirm the fully transformed state of cancer cell exosome-initiated cells

An important step of assessing the tumorigenic property of transformed cells is their ability to form tumors in vivo. To determine whether exosome-initiated transformed cells have the capacity to form tumors when injected subcutaneously into immunocompromised mice, we first isolated and expanded foci cells from the MCA/TPA and Capan-2 exosome/TPA experiments (Figure 3A). The cells from these foci were then injected into NSG (NOD scid gamma) mice at concentrations of 0.1 × 106, 0.5 × 106, and 2.5 × 106 cells to determine the sufficient cell density for tumor formation. Mice were followed for 37 days to measure tumor growth (Figure 3—figure supplement 1A,B). As a control, non-transformed NIH/3T3 cells were injected into mice at the highest concentration (2.5 × 106 cells) (Figure 3—figure supplement 1C). After 37 days, tumor growth was observed in mice injected with cells from the chemically treated MCA/TPA foci and the cancer cell (Capan-2) exosome/TPA treated foci, at all three concentrations tested, whereas no tumor growth was observed in any of the mice injected with non-transformed control cells. In each case, the appearance of the tumors formed correlated with the number of cells injected, as higher concentrations of cells resulted in faster growth and larger tumors (Figure 3—figure supplement 1A,B). Histological analysis confirmed that the tumors are fibrosarcomas, as was expected because cells used in the CTA are NIH/3T3 cells from a mesenchymal origin (Figure 3—figure supplement 1D).

Figure 3 with 2 supplements see all
In vivo studies confirm the fully transformed form of cancer cell exosome-initiated cells.

(A) In vivo assay. NIH/3T3 cells are treated with an initiator and promoter according to the cell transformation assay (42 days total). At the end of the transformation experiment, prior to methanol fixation and staining with crystal violet, foci were isolated, expanded, and established as a transformed cell line. Transformed cells are then subcutaneously injected into mice to monitor for tumor formation. Tumor growth was tracked by measuring tumor volume 2x/week for up to 55 days post injection (Post TCI) or until tumor size exceeded maximum limit. (B) Control mice include injection of untreated NIH/3T3 cells or background foci formed in untreated NIH/3T3 cells. Cells were injected at a concentration of 1 × 106 cells. Injection of untreated cells never resulted in tumor growth; injection of background foci from untreated cells resulted in tumor growth in 6 out of 15 total mice (see figure supplements for additional mice). (C) Results from injections of chemically transformed cells (MCA = initiator/TPA = promoter). Transformed cells were injected at a concentration of 1 × 106 cells. Tumor growth was observed in all mice (n = 5) from three independent experiments (see figure supplements for additional mice). (D) Results from injections of cancer cell exosome-initiated transformed cells; exosomes from four cancer cell lines, Capan-2 (ExC), MIA PaCa-2 (ExM), Panc-1 (ExP), and BxPC-3 (ExBx), were used as an initiator with the promoter TPA. Transformed cells were injected at a concentration of 1 × 106 cells. Tumor growth was observed in all mice (n = 5) for each treatment (see figure supplements for additional mice).

https://doi.org/10.7554/eLife.40226.014

Additional in vivo studies were performed to analyze the tumor forming potential of a variety of foci, including background foci formed in control experiments (Figure 3, Figure 3—figure supplement 1, Figure 3—figure supplement 2). Using a concentration of 1 × 106 cells, we observed that cells collected from MIA PaCa-2 exosome/TPA, Panc-1 exosome/TPA, and BxPC-3 exosome/TPA foci formed tumors in mice at varying size and growth rates (Figure 3D). After injection of untreated background foci, we observed that 6 out of 15 total mice formed tumors. Notably, each of the six tumors (five from the same injected foci) grew later in the time course and at a significantly slower rate compared to initiator and promoter treated transformed foci cells, as has been previously observed (Figure 3B, Figure 3—figure supplement 2A) (Xu and Rubin, 1990).

Proteomic analysis of initiated cells and transformed foci cells

We next used proteomics to analyze molecular changes in cells during the transformation process. Cells that were treated with MCA or Capan-2 exosomes as an initiator were harvested after the 3-day initiator treatment and total protein was analyzed by mass spectrometry. Untreated NIH/3T3 cells were used as a control. No marked global changes in protein content were observed for cells treated with MCA (yellow) or Capan-2 exosomes (blue) when compared to proteins found in untreated NIH/3T3 cells (Figure 4—figure supplement 1, Figure 4—figure supplement 1—source data 1).

Proteomics of transformed cells was also analyzed using mass spectrometry. Three foci from independent wells of both MCA/TPA transformed cells (yellow) and Capan-2 exosome/TPA transformed cells (blue) were compared to untreated NIH/3T3 cells as a control (Figure 4, Figure 4—source data 1). In total >1500 proteins were consistently identified in all three replicates of each sample type (see Materials and methods) (Figure 4). Proteins found in the transformed foci were compared to those found in the untreated control (Figure 4A). To determine whether overlapping proteins found in both control cells and foci are consistently present in each foci, we directly compared these data sets. The ‘common’ proteins found in both the untreated control and the transformed foci were compared using Venn diagrams and showed a high degree of overlap between foci of the same type (Figure 4B). By contrast, the proteins found to be ‘unique’ to each foci and absent in the control cells were compared and showed very little overlap between foci of the same type (Figure 4C). Gene ontology (GO) enrichment analysis was performed on the full set of proteins identified in each of the six foci in order to identify specific molecular functions overrepresented in the protein population (Figure 4D).

Figure 4 with 1 supplement see all
Proteomic profiling of transformed NIH/3T3 cells via mass spectrometry.

(A) Comparison of proteins found in transformed cells resulting from treatment with both an initiator and promoter visualized by Venn diagrams. Three separate foci (F1, F2, F3) from MCA/TPA transformed cells and Capan-2 exosome (ExC)/TPA transformed cells were compared to untreated NIH/3T3 cells (control, gray). Results from three biological replicates were combined for each sample. (B) Comparison of common (overlap) proteins found in each of the six transformed foci samples; common proteins identified in control. (C) Comparison of unique proteins found in each of the six transformed foci samples; unique proteins are absent from control. (D) Gene Ontology enrichment analysis of proteins found in all six foci using PANTHER 14.0. Slim molecular functions identified as overrepresented based on analysis of proteins found in samples.

https://doi.org/10.7554/eLife.40226.017

Exome sequencing reveals mutagenic profiles

To better understand the genetic mechanism of cell transformation, exome sequencing was performed on the same set of transformed cells used for proteomic analysis and tumor mice studies: three MCA/TPA foci (yellow) and three Capan-2/TPA (blue) foci. In addition, three independent control foci were sequenced from untreated cells (gray) and TPA-only treated cells (gray); these are referred to as background foci. The total number of variants found by exome sequencing are visualized by Venn diagrams (Figure 5—figure supplement 1). The total set of variant data was used to generate the principle component analysis (PCA) plot shown in Figure 5. When all 12 foci samples were plotted together, the six transformed foci (MCA/TPA and Capan-2/TPA) appeared to cluster tightly while the six background foci (untreated and TPA-only) showed no clear relationship to one another (Figure 5A). When probed further, PCA of just the six transformed foci showed that there is no clear relationship between these samples (Figure 5B). Additionally, the total set of variant data was analyzed using MutaGene (Goncearenco et al., 2017) to investigate the specific types of nucleotide changes in the each of the 12 foci sequenced. MutaGene is a computational tool used to identify the most likely mutagenic processes associated with a set of variants found from whole exome or genome sequencing. The full mutational profile of each set of variants was decomposed into contributing COSMIC mutational signatures (Figure 6, Figure 6—source data 1). Clustering analysis of these signatures shows that all six transformed foci have similar mutational profiles that vary from the six background foci samples. The top mutational signatures found in each of the six transformed foci are COSMIC Signatures 20 and 15 (Figure 6), both of which are associated with defective DNA mismatch repair (MMR) and microsatellite instability. Foci from untreated cells and TPA-only treated samples did show signatures associated with microsatellite instability, but not as the top contributing signature. Instead, the top COSMIC Signature associated with each of the background foci was found to be COSMIC Signature 3, associated with failure of DNA double-strand break-repair. Considering that mismatch repair was found to be the top-contributing signature of each of the transformed foci, mismatch repair genes were analyzed in more detail for specific mutations (Figure 6—figure supplement 1). Missense mutations were found to be encoded in each of the transformed foci, but none of the foci contained the same mutations. None of the untreated background foci and only one of the TPA-only treated background foci had mutations in the analyzed MMR genes. Considering that mutations in oncogenes are often drivers of cell transformation, we also analyzed the mutational state of the 190 known oncogenes across the 12 foci samples. Results did not indicate a likely driver mutation as very few shared mutations were observed between independent foci of the same type and no common point mutations were found (Figure 5—figure supplement 1, Figure 5—figure supplement 1—source data 1).

Figure 5 with 1 supplement see all
Principle component analysis (PCA) of transformed NIH/3T3 cells.

(A) PCA plot showing relationship between three MCA/TPA transformed foci, three Capan-2 exosome (ExC)/TPA transformed foci, three control foci from TPA-only treated NIH/3T3 cells, and three control foci from untreated NIH/3T3 cells. (B) PCA plot showing relationship between same three MCA/TPA transformed foci and Capan-2 exosome (ExC)/TPA transformed foci in the absence of control samples. Principle component analysis is based on comparison of exome-seq variant data using PLINK's identity-by-state (IBS) estimates.

https://doi.org/10.7554/eLife.40226.021
Figure 6 with 2 supplements see all
Mutational profiles of transformed NIH/3T3 cells.

The 12 samples sequenced via Exome-seq include transformed foci formed from four treatment conditions on NIH/3T3 cells: untreated, TPA-only treated, MCA/TPA treated, and Capan-2 exosome (ExC)/TPA treated. The top COSMIC mutational signatures associated with each sample were identified using MutaGene and clustered based on similarity to generate the heatmap shown. Color range corresponds to the contribution score of each mutational profile.

https://doi.org/10.7554/eLife.40226.024

Discussion

There is a growing interest around exosomes functioning as information carrying cell messengers and as an active participant in cancer development and as a result, numerous studies have been published describing their contribution to cancer progression (Costa-Silva et al., 2015; Melo et al., 2014; Dai et al., 2018; Abdouh et al., 2017; Hamam et al., 2016; Antonyak et al., 2011). In this study, we explored whether pancreatic cancer cell exosomes have a distinct role in the transformation of a normal cell to a malignant form by utilizing a classic two-stage cell transformation assay. We were able to describe a previously uncharacterized function for pancreatic cancer cell exosomes as an initiator in malignant cell transformation. Specifically, cancer cell exosomes were only shown to affect cell transformation when used as an initiator in combination with a promoter. Exosomes had no effect on cells when tested as a promoter or when used as both an initiator and promoter in the two-step CTA. Moreover, exosomes isolated from normal cells were unable to initiate cell transformation. By utilizing proteomics and exome sequencing, we were able to gain more understanding on the process of cell transformation. We propose that a single treatment of cancer cell exosomes on NIH/3T3 cells over 3 days can act in a similar way as the chemical initiator MCA, by randomly incorporating molecular changes into DNA. These random mutations then set the stage for a promoter to induce transformation of cells to form foci that are able to induce tumors when injected into mice.

For studies aiming to elucidate the biological functions of exosomes, it is essential that exosomes are reliably isolated from interfering cellular debris or contaminants. One of the current challenges in the exosome field is the lack of a universally accepted purification method. There are many published protocols on exosome purification from cells but it remains difficult to effectively isolate exosomes from contaminants (Willms et al., 2016; Kowal et al., 2016; Théry et al., 2018; Vergauwen et al., 2017). Major improvements have been made in recent years, one being the recent update to the Minimal Information for Studies of Extracellular Vesicles (MISEV2018) (Théry et al., 2018). In the current study, we aimed to meet the criteria posted in the MISEV2018 in order to produce exosome preparations with minimal co-isolating contaminants. To eliminate the possibility that contaminants might participate in the exosome initiator activity, we further purified exosomes using a sucrose density gradient in order to obtain ‘pure’ exosome fractions (Chiou and Ansel, 2016). Identical experiments with both populations of exosomes demonstrated that both samples could act as an initiator in the CTA, thereby establishing exosomes as the initiating factor in this assay (Figure 2—figure supplement 1B).

The classic two-stage CTA was specifically used as a tool to study cell transformation because it is rigorous, reproducible, and has a well-established phenotype. This assay does not stimulate the entire in vivo neoplastic process, but it can provide essential information about identification of potential carcinogens and their mode of action. We also appreciate that the exosomes used in these assays are derived from human cancer cell lines whereas the ‘normal’ cells used in the CTA are NIH/3T3 cells from murine origin. Previous studies have shown highly conserved molecular functions across human and mouse, including the regulation of cell division, DNA replication, and DNA repair (Monaco et al., 2015). Through this assay, we were able to test the role that exosomes play in NIH/3T3 cell transformation in a well-controlled manner.

Recent studies demonstrate that cancer cell exosomes can in fact participate in cell transformation, however the details on how these exosomes contribute to transformation remained elusive. For example, Dai et al. showed that exosomes participated in the transformation process of normal cells after cells were treated with arsenite (Dai et al., 2018). Another study described a transfer of malignant traits to BRCA1 deficient human fibroblast when treated with either cancer patient sera or isolated cancer cell exosomes, leading to malignant transformation (Abdouh et al., 2017; Hamam et al., 2016). In addition, exosomes in breast cancer patient sera have been shown to promote nontumorigenic epithelial cells to form tumors in a Dicer-dependent manner (Melo et al., 2014). In each of the aforementioned studies, non-transformed cells were either pre-exposed to transforming agents or treated with cancer patient sera or medium from cultured cancer cells. Interestingly, Antonyak et al. has demonstrated that sustained treatment of NIH/3T3 cells with breast cancer or glioma cell-derived microvesicles has the ability to induce transformation of recipient cells (Antonyak et al., 2011). Contrastingly, in our studies, when cells were treated only with pancreatic cancer cell exosomes, increased cell transformation was not observed. While the specific details and effects on cell transformation attributed to exosomes vary between the studies, both demonstrate that exosomes are indeed playing a role in cell transformation.

Collectively, we observe that exosomes can function as an initiator in malignant cell transformation. Our results show that this activity is independent of the KRAS mutational state of exosome producing cells (Figure 2C). Boiling exosomes destroys the membrane structures and releases the vesicle content into the media, demonstrating that intact vesicles are needed to induce cell transformation (Figure 2—figure supplement 1A). Additionally, exosomes were found to work as a classic initiator and thus function with multiple promoters (TPA or CdCl2) (Figure 2D). Finally, when cells isolated from MCA/TPA and cancer cell exosome/TPA foci were injected subcutaneously into immunocompromised mice, tumors formed (Figure 3, Figure 3—figure supplements 1 and 2).

We propose that, like MCA, cancer cell exosomes can cause random molecular changes to mediate the initiator step in cell transformation. Then, treatment with the promoter TPA forces cells to proliferate repeatedly and drives them to be fully transformed. The biochemistry associated with MCA-driven transformation is complex and diverse; it is known to produce bulky carcinogen-DNA adducts, which are associated with G·C → T·A transversions and thus introduces mutagenic and carcinogenic properties into DNA (Malins et al., 2004). According to Malins et al., MCA creates an identifiable tumor phenotype long before the appearance of tumors and causes changes in DNA structure that could be expected to influence gene expression and further translation (Malins et al., 2004). Although we observe no major changes in the protein content in initiated cells based on qualitative proteomic analysis, it is logical that changes could not be feasibly detected because they are occurring in specific proteins among specific cells in the total population. Additionally, it is known that exosomes do not uniformly contain the same pool of proteins, lipids, metabolites, or microRNAs, but rather each exosome contains a unique repertoire of biological molecules thus possibly causing a variety of changes in the cells (Willms et al., 2016; Kowal et al., 2016; Smith et al., 2015). Therefore, potentially both MCA and cancer cell exosomes are causing random molecular changes across the population of treated NIH/3T3 cells that cannot be detected amongst the overall protein composition of cells by mass spectrometry. Ultimately, more studies are needed to elucidate the molecular mechanism of this cancer cell exosome-mediated initiation event.

To further investigate the effects of cancer cell exosome-initiated transformation on cells, we probed the transformed foci that are formed during the two-stage cell transformation by proteomics and exome sequencing. Comparing the proteins found in MCA/TPA foci and Capan-2 exosome/TPA foci (transformed foci) revealed a set of unique proteins not identified in the control NIH/3T3 cells (Figure 4A). Venn diagrams comparing the proteins found in each foci reveal that these unique proteins vary between foci of the same type (Figure 4B and C). GO enrichment analysis of the transformed foci highlight this diversity, as a minority of molecular functions are found to be enriched in all the transformed foci while the majority of enriched functions vary between foci (Figure 4D). These studies, in contrast to the proteomic studies on the initiator treated cells, show major changes in the proteomic profiles of transformed cells (Figure 4D).

We performed exome sequencing analysis on twelve foci samples in total: three untreated control foci and three TPA-only treated foci (referred to as background foci); three MCA/TPA foci and three Capan-2 exosome/TPA foci (referred to as transformed foci). As expected, we observed diversity in mutational variants for each of these pools of cells because each foci was derived from independent mutagenic events (Figure 5—figure supplement 1). PCA demonstrated that the transformed foci appear to cluster together while the background foci show no clear relationship to one another. Using PCA for just the transformed foci, we do observe heterogeneity between each sample, supporting the diversity in variants observed from sequencing (Figure 5, Figure 5—figure supplement 1). We then used MutaGene analysis with the full set of variants for each foci to generate the mutational profile for each. The data in Figure 6 represents the relative contribution of the pan-cancer derived COSMIC signatures to the mutational profiles of each foci. Although all 12 samples are divergent at the molecular level, there is a distinct difference between the top contributing mutational signatures found in the six transformed foci as opposed to the six background foci (Figure 6).

The top signatures identified in the transformed foci (COSMIC Signatures 20 and 15), are associated with defective DNA mismatch repair and microsatellite instability (Figure 6). Consistent with this observation, analysis of MMR genes for specific mutations revealed mismatch repair gene mutations in these six transformed foci. Furthermore, based on in silico prediction analysis we observed that the mutations detected in mismatch repair genes may be deleterious or damaging for the activity of the mismatch repair proteins (Figure 6—figure supplements 1 and 2). Concurrently, sequencing data shows that the wild-type allele for each MMR gene is still present in these six foci. While MMR genes usually comply with Knudson’s two-hit hypothesis for tumor suppressor genes, the presence of the wild-type copy of a MMR gene in somatic cells is not always sufficient for a normal function. Haploinsufficiency may be function-specific; for example, it has been demonstrated that responding to DNA damage requires a higher dosage of MMR protein than the repair function (Peltomäki, 2016). Regardless, the dysfunction of the MMR system is supported by the presence of the COSMIC profiles 20 and 15. Interestingly, none of the foci contain the same mutations in the MMR genes, further supporting the proposal that transformed cells originated from distinct mutagenic events. By contrast, only a single sample from the background foci was observed to have mutations (two) in MMR genes. In addition, the main driver found for all six background foci was COSMIC Signature 3. This profile associates with a failure in DNA double-strand break-repair that compromises genomic integrity and likely contributes to the high number of variants found in these samples (Figure 5—figure supplement 1).

Analysis of the top signatures for all six transformed foci shows that mutations in mismatch repair machinery are most likely what drives the transformation of these cells. Although the initiators (MCA and Capan-2 exosomes) varied for these six transformed foci, all were derived from NIH/3T3 cells and used TPA as a promoter for cell transformation. Analysis of the top signatures and perusal of mutated genes for the background foci demonstrates drivers other than MMR genes were used for the transformation of the cells. These observations support the proposal that when TPA is used as a promoter (to drive proliferation) in combination with a functional initiator (an entity that creates a population of randomly mutated cells), the molecular path of cells towards tumorigenesis may not be random. Rather, TPA may drive a path in a subset of randomly mutated cells that accommodates a course towards COSMIC signatures 20 or 15. Furthermore, future studies might reveal that another promoter acting on initiator treated cells might lead cells towards another common endpoint with regards to the 30 different signature profiles.

A recent study by Felsentein et al. described an interesting finding in which a portion of co-occurring IPMNs in PDAC patients appeared to be genetically unrelated, meaning they shared no mutations in the assayed genes (Felsenstein et al., 2018). This elicits a fascinating question as to whether cells in a primary tumor could be derived from independent transformation events as opposed to exclusive clonal events (Felsenstein et al., 2018). Potentially, exosomes secreted by the primary tumor could orchestrate such events. Additional studies towards understanding the cancer cell exosome-mediated initiation are needed to address such questions, specifically, investigating the initiation capacity of cancer cell exosomes in more relevant cells like human epithelial cells.

In conclusion, we observe that cancer cell exosomes have the capacity to act as a classic initiator in the 2-stage CTA by incorporating random changes to NIH/3T3 cells over three days to mediate the first step in cell transformation (Figure 7). We observe that exosomes from pancreatic cancer cell lines, independently from the KRAS mutation status, can initiate the transformation of NIH/3T3 cells, while exosomes from normal pancreatic cells do not possess this ability. Future studies expanding on how cancer cell exosomes can uniquely function as an initiator are under way. Importantly, these observations provide insight into the molecular role of cancer cell exosomes in cell transformation and how this activity might contribute to the dynamic conversation between normal and cancer cells.

Schematic model of exosome mediated transformation.

Exosomes secreted by cancer cells are taken up by normal NIH/3T3 cells and have the capacity to act as an initiator by incorporating random changes into the recipient cell genome. These initiated cells, when exposed to a promoter, can be induced by further alterations to a transformed state that has the ability to grow into a malignant tumor.

https://doi.org/10.7554/eLife.40226.028

Materials and methods

Key resources table
Reagent type
or resource
DesignationSource or
reference
IdentifiersAdditional
information
Chemical12-O-Tetradecanoylphorbol-13-acetate (TPA)Cell Signaling Technology4174
ChemicalMethylcholanthrene (MCA)Sigma-Aldrich213942–100 MG
ChemicalCadmium Chloride (CdCl2)Sigma-Aldrich655198–5G
ChemicalDimethyl sulfoxide (DMSO)Sigma-AldrichD2650−5 × 5 ML
Cell line (Mus musculus)NIH/3T3ATCCRRID:CVCL_0594
Cell line (Homo-sapiens)Dermal fibroblast (normal, Adult)ATCCPCS-201–012
Cell line (Homo-sapiens)Capan-2ATCCRRID:CVCL_0026
Cell line (Homo-sapiens)PANC-1ATCCRRID:CVCL_0480
Cell line (Homo-sapiens)MIA PaCa-2ATCCRRID:CVCL_0428
Cell line (Homo-sapiens)BxPC-3ATCCRRID:CVCL_0186
Cell line (Homo-sapiens)HPDE (H6C7)KerafastRRID:CVCL_0P38
AntibodyAnti-ALIX (3A9) (mouse monoclonal)AbcamAbcam, Cat#A2228, RRID:AB_10899268(1:500)
Antibodyα-actinin (H-2) (mouse monoclonal)Santa Cruz BiotechnologySanta Cruz Biotechnology, Cat#sc-17829, RRID:AB_626633(1:1000)
AntibodyAnti-β-actin (AC-74) (mouse monoclonal)Sigma-AldrichSigma-Aldrich, Cat#A2228, RRID:AB_476697(1:5000)
AntibodyCalnexin (C5C9) (rabbit monoclonal)Cell Signaling TechnologyCell Signaling Technology Cat# 2679, RRID:AB_2228381(1:1000)
AntibodyCD63 (rabbit polyclonal)ProteintechProteintech, Cat#25682–1-AP, RRID:AB_2783831(1:1000)
AntibodyHSP90α/β (F8) (mouse monoclonal)Santa Cruz BiotechnologySanta Cruz Biotechnology, Cat#sc-13119, RRID:AB_675659(1:1000)
AntibodyTSG101 (4A10) (mouse monoclonal)Thermo Fisher ScientificThermo Fisher Scientific Cat# MA1-23296, RRID:AB_2208088(1:500)
Software, algorithmGraphPad Prism 8GraphPadRRID:SCR_002798
Software, algorithmProteome
Discoverer 2.1
Thermo ScientificThermo Fisher Scientific, RRID:SCR_014477
Software, algorithmFASTQC v0.11.5Babraham BioinformaticsRRID:SCR_014583
Software, algorithmTrim GaloreBabraham BioinformaticsRRID:SCR_011847
Software, algorithmBurrows-Wheeler Aligner (BWA)Burrows-Wheeler AlignerRRID:SCR_010910
Software, algorithmStrelka2IlluminaRRID:SCR_005109
Software, algorithmVCFtoolsVCFtoolsRRID:SCR_001235
Software, algorithmSnpEffSnpEffRRID:SCR_005191
Software, algorithmPLINKPurcell et al., 2007RRID:SCR_001757
Software, algorithmPythonPython Software FoundationRRID:SCR_008394
Software, algorithmMutaGeneMutaGeneRRID:SCR_016574
Software, algorithmPROVEANJ.Craig Venter InstituteRRID:SCR_002182

Chemicals

12-O-Tetradecanoylphorbol-13-acetate (TPA) (Cell Signaling Technology), Methylcholanthrene (MCA) (Sigma-Aldrich), Cadmium Chloride (CdCl2) (Sigma-Aldrich), Dimethyl sulfoxide (DMSO) (Sigma-Aldrich). Each compound was dissolved in DMSO and preserved at −20°C.

Cells and culture conditions

Request a detailed protocol

Mouse embryo cell line, NIH/3T3, human primary dermal fibroblast, human pancreatic cancer cell lines: CAPAN-2, Panc-1, MIA PaCa-2, and BxPC-3, were purchased from American Type Culture Collection (ATCC, Manassas, VA). Immortalized human pancreatic duct epithelial cell line, HPDE, was from Kerafast (Kerafast, Boston, MA). Capan-2, MIA PaCa-2, and Panc-1 were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Millipore Sigma) supplemented with 10% (v/v) fetal bovine serum (FBS, Millipore Sigma) and 1% (v/v) antibiotics solution (Penicillin-Streptomycin, Millipore Sigma). BxPC-3 cells were maintained in minimum essential media (MEM) (Fisher) supplemented with 10% (v/v) fetal bovine serum (FBS) and 1% antibiotics solution (Penicillin-Streptomycin, Millipore Sigma). HPDE cells were maintained in Keratinocyte Serum-Free Media (KSFM, Invitrogen) with KSFM Supplements including epithermal growth factor (EGF) and bovine pituitary extract (BPE) (Invitrogen). Primary dermal fibroblast were maintained in Fibroblast Basal Medium (ATCC PCS-201–030) with Fibroblast Growth Kit-Serum-free (ATCC PCS-201–040) supplements included. NIH/3T3 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Millipore Sigma) supplemented with 10% (v/v) Bovine Calf serum (CS, Gemini) and 1% antibiotics solution (Penicillin-Streptomycin, Millipore Sigma). All cell lines were cultured at 37°C in a humidified atmosphere of 5% CO2. Each cell line was tested free from mycoplasma. NIH/3T3 cells were used below passage 4 (p<4), primary dermal fibroblast (p<8), HPDE cells (p<8), and carcinoma cell lines (p<20).

Experimental design

Exosome isolation, subcellular fractionation, and TCA precipitation

Request a detailed protocol
Exosome isolation

Exosomes were isolated using a previously described combined ultrafiltration-ultracentrifugation protocol (Adamczyk et al., 2011). In detail, pancreatic cancer cells (CAPAN-2, Panc-1, MIA PaCa-2, and BxPC-3) and normal human cells (HPDE and human primary dermal fibroblasts) were grown in ten 225 cm3 flasks in standard medium until they reached a confluency of approximately 70–80% (~3.5×108 cells). The carcinoma cell lines were then washed twice with medium and incubated in plain, serum-free medium for 72 hr. For HPDE cells and human primary dermal fibroblasts, phosphate-buffered saline (PBS) was used for washing and plain Keratinocyte SFM medium without supplements was used for exosome production for 72 hr. This protocol did not measurably increase the rate of cell death as determined by trypan blue exclusion, which showed over 93% live cell counts after 72 hr incubation in conditioned media. Next, the conditioned media (approximately 450 mL) from serum-free cell cultures were cooled down on ice, centrifuged (200xg, 10 min), and passed through 0.2 μm pore filters to remove cells, cell debris, and vesicles sized smaller than 220 nm. An inhibitor cocktail was added to protect the proteins from proteolytic digestion (PMSF and inhibitor cocktail complete Roche, Mannheim, Germany). Enrichment of exosomes was accomplished by subsequent ultrafiltration with Amicon Ultra 100K (4000xg, 25 min, 4°C), followed by ultracentrifugation at 120,000 g for 90 min at 4°C. The exosome pellet was washed in PBS followed by another ultracentrifugation at 120,000 g for 90 min at 4°C. The final exosome pellet was resuspended in PBS and validated by characterization via mass spectrometry analysis of proteins, western blot analysis, electron microscopy analysis (TEM), and nanoparticle tracking analysis (NTA) (Lötvall et al., 2014; Witwer et al., 2017; Théry et al., 2018). Each exosome pellet, after final resuspension into PBS, was divided into 5 µl aliquots, stored at −80°C, and thawed right before use. The protein concentration of the exosome fraction was measured after each exosome isolation using CBQCA protein quantitation kit (Invitrogen). Exosomes were used in the cell transformation assay experiments at a concentration of 80 ng/mL (equivalent to 7 × 107 particles/mL) based on dose response studies.

Sucrose gradient separation

In addition to the exosome isolation protocol described above, exosomes were further purified using floatation into a sucrose density gradient (Chiou and Ansel, 2016). In detail, sucrose gradients were built manually as described in reference five by first preparing 12 sucrose stock fractions in PBS with sucrose concentrations ranging from 10–90%. Half of the exosome pellet from the isolation protocol described above was resuspended in 50 µl PBS with 1 ml of 90% sucrose stock solution and loaded at the bottom of a 13.2 mL ultra-clear Beckman ultracentrifuge tube. The gradient was layered by sequentially pouring 1 mL of each of the remaining 11 solutions in order from highest to lowest sucrose concentration. Tubes were centrifuged for 16 hr at 4°C at 100,000 g (24200 rpm) in a TH-641 rotor. At completion, six 2 mL fractions were collected from each of the tubes. Next, 9 mL of PBS was added to each of the 2 mL fractions and centrifuged at 4°C at 100,000 g for 1 hr. The supernatant was carefully aspirated before the pellet was resuspended in 50 µl of PBS and validated by mass spectrometry, western blot, and nanoparticle tracking analysis (NTA). The protein concentration of the fractions was measured using CBQCA protein quantitation kit (Invitrogen). Fraction 3, shown to contain exosomal marker proteins, was used in the cell transformation assay experiment at 80 ng/mL.

Subcellular fractionation

Cells from exosome preparations were harvested from one 225 cm3 flask, after the conditioned media for exosome isolation was collected, using 0.25% trypsin-EDTA (Sigma-Aldrich) treatment, followed by lysing according to previously published protocol (Casey et al., 2018). In brief, cells were suspended in HNMEK lysis buffer (20 mM HEPES pH 7.4, 50 mM NaCl, 2 mM MgCl2, 2 mM EDTA, 10 mM KCl, 50 nM EGTA, protease inhibitors) and lysed using a Dounce homogenizer. Lysates were centrifuged at 500 g for 10 min at 4°C to remove nuclei and cellular debris. The supernatant was collected and centrifuged at 10,000 g for 10 min to pellet ER and mitochondrial membranes (P2 fraction). Cytoplasmic S2 fraction was collected and kept for further analysis. P2 fraction was washed once in HNMEK lysis buffer, centrifuged at 10,000 g for 10 min, and the pellet was resuspended in RIPA lysis buffer (50 mM Tris pH 8, 150 mM NaCl, 5 mM EDTA pH8, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 1 mM PMSF, protease inhibitors). Protein concentration was measured from each fraction using the Bradford Protein assay (Bio-Rad).

TCA precipitation

The last fraction collected from the exosome isolation procedure was the flow through media (M). This flow through consists of 20% starting conditioned media after the ultrafiltration step, and was kept for deoxycholate-trichloroacetic acid precipitation (DOC-TCA) of the proteins. This medium represents a sample that should not contain any exosomal proteins when analyzed for exosome markers using western blot analysis. In brief, 0.15% DOC was added to media samples (1:10 ratio) and incubated on ice for 15 min followed by addition of TCA to 8% final concentration and incubated overnight at 4°C. Precipitated proteins were pelleted by centrifuging at 18,000 rpm and washed twice with cold acetone, prior to re‐suspension in 10 mM Tris–HCl pH = 8.0. The protein concentration was measured using CBQCA protein quantitation kit (Invitrogen).

Cell transformation assay

Request a detailed protocol

The two-stage cell transformation assay was carried out according to the protocol described by Sakai and Sato (1989) with small modifications. In detail, the frozen stock of NIH/3T3 cells were thawed and cultured. Actively growing cells with passage number <4 were seeded for the transformation assay at a density of 2.5 × 103 cells/well in a 6-well plate with 2 mL of culture medium. Two days after seeding, media was replaced with media containing an initiator, either MCA (0.5 µg/mL), exosomes (80 ng/mL), or 0.02% DMSO, and cells were grown for 3 days. Next, the medium was replaced with fresh medium and cells were grown for an additional 2 days. Cultures were then treated with a medium containing a promoter, either TPA (300 ng/mL), exosomes (80 ng/mL), CdCl2 (120 ng/mL), or 0.2% DMSO, for 2 weeks. The cells were subsequently cultured in normal medium for 3 weeks. The medium was changed every other day during the promoter treatments and twice a week for the last 3 weeks of the experiment. The cells were fixed with methanol and stained with crystal violet for focus scoring. Each test chemical was dissolved in DMSO. The concentration of vehicle was below 0.2%, which did not affect the induction of transformed foci.

Foci scoring

Request a detailed protocol

The scoring of transformed foci was carried out according to established criteria on focus scoring (Sasaki et al., 2012). Different categories of foci that can be observed are Type I, Type II, and Type III. Only Type III foci are scored as malignantly transformed and were counted as positive in this study. Foci were assessed for the following characteristics: deep basophilic staining, spindle-shaped cells, multilayer growth (piling up of cells), random orientation at the edge of the focus, and invasive growth into the background monolayer; each characteristic needed to be present for Type III classification. To ensure accurate scoring, foci scoring was performed in a double-blinded manner by two researchers.

Tumorigenicity of transformed cells

Request a detailed protocol

All animals were housed in a pathogen-free facility with 24 hr access to food and water. Experiments were approved by, and conducted in accordance with, an IACUC approved protocol at UT Southwestern. Six-to-8-week-old female NOD/SCID mice were obtained from an on-campus supplier.

At the end of the transformation experiment, prior to methanol fixation and staining with crystal violet, MCA/TPA-treated and cancer cell exosome/TPA-treated cells that had formed type III foci were isolated, expanded, and established as a transformed cell line. To assess their tumorigenic property, cells were injected subcutaneously into mice. To determine the sufficient cell density for tumor formation, cells were first isolated from MCA/TPA-treated and Capan-2 exosome (ExC)/TPA-treated experiments and injected into mice at three different cell concentrations (0.1 × 106, 0.5 × 106, and 2.5 × 106 cells). Untreated NIH/3T3 cells were injected into mice as a control at the highest concentration used (2.5 × 106 cells). Every group consisted of n = 5 mice. Mice were observed for tumor formation by palpating twice a week and recording the weight of the mice and size of the tumor. The experiment was terminated 37 days post injection, when the tumor size in some of the animals had reached maximum allowable diameter. Histological analyses confirmed that the tumors are fibrosarcomas, as expected from the transformed cells of mesenchymal origin. Additional in vivo experiments include subcutaneous injection of mice (n = 5/group) with 1 × 106 cells isolated after treatment with MIA PaCa-2 (ExM)/TPA, Panc-1 (ExP)/TPA, or BxPC-3 (ExBx)/TPA as well as injection of three background foci formed from untreated NIH/3T3 cells. Tumor growth was tracked in the same manner up to 55 days post injection or until tumor growth exceeded maximum allowable size.

Initiation assay

Request a detailed protocol

NIH/3T3 cells were plated on a six-well plate at a density of 2.5 × 103 cells/well with 2 mL of culture medium. Two days after seeding, media was replaced with fresh complete media or media containing an initiator; either 0.5 µg/mL MCA or 80 ng/mL Capan-2 exosomes, and cells were grown for 3 days, followed by 2 days of recovery (two wells for each condition). Next, cells were harvested and lysed with RIPA lysis buffer (50 mM Tris pH 8, 150 mM NaCl, 5 mM EDTA pH 8, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 1 mM PMSF, protease inhibitors) by incubating cells on ice for 30 min and vortexing twice during the incubation. At the end of incubation, lysates were centrifuged at 10,000 g to pellet cell debris. Protein concentration was measured using the Bradford Protein assay (Bio-Rad). Experiments were performed in triplicate for each treatment condition. Total protein composition (26 µg of protein/sample) was analyzed by mass spectrometry.

Preparation of samples for mass spectrometry and exome sequencing analysis

Request a detailed protocol
Proteins

For analysis of exosomal proteins by mass spectrometry, exosome samples (equivalent of approximately 15 μg of protein) isolated from each cell line were thawed from −80°C storage and 10 μl of 5 × protein sample buffer was added. Samples were boiled for 5 min and loaded on TGX stain‐free gels (Bio‐Rad) and run 10 mm into the top of an SDS–PAGE gel. Gel bands containing proteins were excised for mass spectrometry analysis.

For analysis of the total protein composition of untreated NIH/3T3 cells, transformed cells, and cells from the initiation assay, cells were lysed in RIPA lysis buffer and protein concentration was measured using the Bradford Protein assay (Bio-Rad). Equal amounts of proteins (approximately 26 µg of protein from the initiation assay samples and 35 µg of protein from the transformed cell samples) from three biological replicates were taken and 5 × protein sample buffer was added. Samples were boiled for 5 min and loaded on TGX stain‐free gels (Bio‐Rad) and run 10 mm into the top of an SDS–PAGE gel. Gel bands containing proteins were excised for mass spectrometry analysis.

Genomic DNA

DNA for exome sequencing was extracted from untreated NIH/3T3 cells, three background foci from untreated NIH/3T3 cells, three background foci from TPA-only treated NIH/3T3 cells, three foci from MCA/TPA treated NIH/3T3 cells, and three foci from ExC/TPA treated NIH/3T3 cells using Quick-gDNA MiniPrep Kit (Genesee Scientific) according to the manufacturer’s protocol. Quality and concentration of the DNA was measured with BioAnalyzer. Exome libraries were prepared using IDT xGen Exome Research Panels and sequenced on Illumina HiSeq 4000 at 100x coverage.

Data analysis

Validation of exosome isolation

Request a detailed protocol
Transmission electron microscopy (TEM)

Electron microscopy was used to characterize vesicles pelleted by the ultrafiltration-ultracentrifugation isolation method described above and to provide information on the size of the vesicles. TEM negative staining was performed on the aliquots stored at −80°C. 10 µl of exosome suspension in PBS was placed onto carbon-coated grids (200mesh) for 1 min and negatively stained with 2% uranyl acetate solution for 1 min. Grids were visualized at 13000x to 68000x in a FEI Tecnai G2 Spirit transmission electron microscope at 120kV. Separate images were taken to provide a wide field encompassing multiple vesicles or to provide close-up images of single vesicles.

Nanoparticle tracking analysis

Nanoparticle tracking analysis (NTA) was performed using NanoSight Version 2.3 on crude exosomes and exosomes further purified by sucrose density gradient (Fraction three only). Finite Track Length Analysis (FTLA) was used for size determination. In each case, average vesicle size per concentration was determined from three measurements of a single exosome preparation.

Mass spectrometry

Mass spectrometry analysis was performed on crude exosomes derived from Capan-2 cells or purified exosomes (Fraction 3) derived from Capan-2 cells. Figure 1—figure supplement 2 shows a list of the top 20 most commonly found proteins in exosomes according to the Exocarta database (Kowal et al., 2016). All twenty proteins were identified by mass spectrometry in both populations of Capan-2 cell exosomes. Complete proteins lists from crude exosomes and Fraction three exosomes are included as Figure 1—figure supplement 2—source datas 1 and 2.

Antibodies and western blot

Alix antibody (cat. no. ab117600, Abcam), α-actinin antibody (H-2, cat. no. sc-17829, Santa-Cruz), β-actin antibody (Clone AC-74, cat. no. A2228, Sigma-Aldrich), Calnexin antibody (Clone C5C9, cat. no. 2679, Cell Signaling), CD63 antibody (cat. no. 25682–1-AP, Proteintech), HSP90α/β (F-8, cat. no. sc-13119, Santa-Cruz), TSG101 antibody (Clone 4A10, cat. no. MA1-23296, Thermo Fisher).

Western blots were used to examine the presence of common exosomal proteins in cellular fractions and sucrose gradient purified exosomes. Using Capan-2 cells as a representative example, equivalent micrograms of proteins from ER and mitochondrial (P2), cytoplasmic (S2), media (M), and exosome (Ex) fractions, prepared from different steps during exosome isolation as described above in section 1.1, were separated by SDS-PAGE and transferred to nitrocellulose membranes. Development was performed using Pierce ECL 2 Western blotting substrate (Thermo Fisher Scientific) and radiographic films (Lightlab).

Cytotoxicity and colony formation efficiency of cadmium chloride (CdCl2)

Request a detailed protocol

Cytotoxicity and colony formation efficiency assays were used to determine a suitable concentration of CdCl2 for use as a promoter in cell transformation assay. Cytotoxicity of CdCl2 was determined using a protocol as described by Umeda et al. (1989) and Fang et al. (2001) with some modifications. Experiments were repeated twice to confirm reproducibility. In brief, toxicity was tested by plating 2 × 104 cells/mL into 6-wells and culturing for 24 hr. Next, cells were treated with 40, 120, 240, or 360 ng/mL CdCl2, with 3-wells/each concentration. After a 4 day cultivation, the cell number of each well was determined after trypsin treatment using trypan blue exclusion method by counting the number of live cells to the number of dead cells. Cell viability remained high (>89%) for the first three concentrations (40, 120, 240 ng/mL) but dropped to an average of 56% in the highest concentration (360 ng/mL).

For the colony formation efficiency assay, cells were plated at 200 cells/2 mL into 6-wells and cultured for 24 hr. Next, cells were treated with 40, 120, 240, or 360 ng/mL CdCl2, 3-wells/each concentration, for 10 days. Media was changed on day 5. The cells were fixed with methanol and stained with crystal violet for counting the number of colonies. Only colonies comprising >50 cells were scored. Relative colony formation efficiency was calculated as (%) = (number of test colonies/number of control colonies) x 100 (Fang et al., 2001). At the two highest concentrations (240, 360 ng/mL) colony formation was reduced to 4% when compared to control, untreated cells. At 120 ng/mL colony formation was inhibited by 50% and cell viability maintained above 89%. The lowest concentration, 40 ng/mL, did not have any effect on the colony formation efficiency when compared to control colony formation. The highest CdCl2 concentration to inhibit colony formation by 50% while retaining cell viability was 120 ng/mL and was therefore used in cell transformation assays.

Dose response studies

Request a detailed protocol

Dose response experiments were performed to determine a suitable concentration of exosomes for use as an initiator in cell transformation assays. Protein concentration was used as a normalization strategy. Transformation assays with different protein concentrations of exosomes were repeated twice to confirm reproducibility. Cells were treated with 0.08, 0.8, 8, 24, 80, 240, 800 or 2400 ng/mL of exosomes as an initiator as described in detail in section 1.2; 6-wells per concentration were used. At the end of the cell transformation assay, cells were fixed with methanol and stained with crystal violet. Focus scoring were performed according to the method described in section 1.3. Equal transformation induction activity was observed for all concentrations except the two lowest ones, 0.08 ng/mL and 0.8 ng/mL. A standard concentration of 80 ng/mL (equivalent to 7 × 107 particles/mL) of proteins were used in each transformation assay.

Statistical analysis of foci scoring

Request a detailed protocol

Statistical analysis of foci scoring was performed by two-tailed unpaired t test with Welch's correction.

Proteomic analysis by mass spectrometry

Request a detailed protocol

Sample preparation for mass spectrometry analysis included the excision of proteins from polyacrylamide gels via SDS-PAGE and Coomassie blue dye staining. Protein samples were reduced and alkylated using DTT and iodoacetamide, respectively. Samples were digested overnight using trypsin (37°C) and resulting peptides were de-salted using solid phase extraction (SPE). LC-MS/MS experiments were performed on a Thermo Scientific EASY-nLC 1200 liquid chromatography system coupled to a Thermo Scientific Orbitrap Fusion Lumos mass spectrometer. To generate MS/MS spectra, MS1 spectra were first acquired in the Orbitrap mass analyzer (resolution 120,000). Peptide precursor ions were then isolated and fragmented using high-energy collision-induced dissociation (HCD). The resulting MS/MS fragmentation spectra were acquired in the ion trap. MS/MS spectral data from exosome samples was searched using Proteome Discoverer 2.1 software (Thermo Scientific) against entries included in either the Human Uniprot protein database (173,060 entries) (exosome samples) or the Mus musculus (Mouse) Uniprot protein database (86,520 entries) (NIH/3T3 cell samples). Search parameters included Carbamidomethylation of cysteine residues (+57.021 Da) as a static modification and oxidation of methionine (+15.995 Da) and acetylation of peptide N-termini (+42.011 Da) as dynamic modifications. The precursor ion mass tolerance was set to 10 ppm and the product ion mass tolerance was set to 0.6 Da for all searches. Peptide spectral matches were adjusted to a 1% false discovery rate (FDR) and proteins were filtered to a 1% FDR.

For exosome samples, only a single run was analyzed for the presence of exosomal marker proteins. For transformed NIH/3T3 cells (Figure 4) and initiated NIH/3T3 cells (Figure 4—figure supplement 1) each sample was run in biological triplicate. Additional filtering was applied to protein datasets from transformed NIH/3T3 cells (data contained in Figure 4) to only compare proteins identified in all three biological replicates. Complete proteins lists from crude exosomes and Fraction three exosomes are included as Figure 1—figure supplement 2—source datas 1 and 2. Complete protein lists from initiated NIH/3T3 cells are included as Figure 4—figure supplement 1—source data 1. Complete protein lists from transformed NIH/3T3 cells are included as Figure 4—source data 1.

Gene ontology (GO) enrichment analysis was performed using PANTHER 14.0 on the set of proteins identified in each of the six transformed foci shown in Figure 4 (three MCA/TPA foci and three ExC/TPA foci). Each protein list contained proteins identified in all three biological replicates for each foci. Protein lists were searched against the reference Mus musculus protein database to identify overrepresented molecular functions (Ashburner et al., 2000; The Gene Ontology Consortium, 2019; Mi et al., 2017).

Exome sequencing analysis

Request a detailed protocol
Variant calling

Raw reads were quality controlled and filtered using FASTQC v0.11.5 (Andrews, 2016) and Trim Galore v0.4.1 (Krueger, 2015) using default settings. Reads were mapped to the mouse reference genome (GRCm38) using BWA-MEM v0.7.12 (Li and Durbin, 2009). Somatic mutations were called using Strelka2 v2.9.0 against the original NIH/3T3 cells (Kim, 2018). Mutations were filtered for quality using VCFtools v0.1.14 that meet PASS criteria based on Empirical Variant Score and minimum read depth of DP >10. Comparison between similarly treated foci were compared using ‘vcf-compare’ using VCFtools v0.1.14 and Venn diagrams comparing all variants were drawn using Python v3.6.8.

Annotation and foci comparison

Mutations were annotated using SnpEff v4.3q (PMID:22728672) for loss-of-function or missense mutation. Mutation-based clustering analysis was performed on full variant data set for each foci using pairwise identity-by-state function in PLINK v.1.90b4 (Purcell et al., 2007) and visualized using the first two principal components using Python v3.6.8 to compare similarity between foci. Annotations were filtered for loss-of-function or missense mutation in order to generate the table of non-synonymous variants found in mismatch repair associated genes in Figure 6—figure supplement 1.

Generating mutational signatures and heatmap

Mutational signatures were derived from the full set of variants for individual samples using MutaGene (Goncearenco et al., 2017). To identify overlapping signatures between samples, we performed hierarchical clustering by calculating the Euclidean distance using clustermap from seaborn v0.7.1[@doi:10.5281/zenodo.54844].

In silico analysis

PROVEAN v1.1.3. (Protein Variation Effect Analyze, http://provean.jcvi.org/index.php) genome variant software was used to predict the potential impact of the identified missense variants on protein function in the mismatch repair associated genes. This tool provides PROVEAN and SIFT predictions for a list of genome variants.

Data and materials availability

Request a detailed protocol

All data are available in the main text or the supplementary materials.

References

  1. 1
  2. 2
  3. 3
    Focus formation: a cell-based assay to determine the oncogenic potential of a gene
    1. A Alvarez
    2. GA Barisone
    3. E Diaz
    (2014)
    Journal of Visualized Experiments, 10.3791/51742, 25590282.
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
    Improved exosome isolation by sucrose gradient fractionation of ultracentrifuged crude exosome pellets
    1. NT Chiou
    2. KM Ansel
    (2016)
    Protocol Exchange, 10.1038/protex.2016.057.
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
    Trim Galore, version 0.4.1
    1. F Krueger
    (2015)
    Babraham Bioinformatics.
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
    BALB/c 3t3 cell transformation assays for the assessment of chemical carcinogenicity
    1. A Sakai
    (2007)
    Alternatives to Animal Testing and Experimentation 14:367–373.
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
    Minimal information for studies of extracellular vesicles 2018 (MISEV2018): a position statement of the international society for extracellular vesicles and update of the MISEV2014 guidelines
    1. C Théry
    2. KW Witwer
    3. E Aikawa
    4. MJ Alcaraz
    5. JD Anderson
    6. R Andriantsitohaina
    7. A Antoniou
    8. T Arab
    9. F Archer
    10. GK Atkin-Smith
    11. DC Ayre
    12. JM Bach
    13. D Bachurski
    14. H Baharvand
    15. L Balaj
    16. S Baldacchino
    17. NN Bauer
    18. AA Baxter
    19. M Bebawy
    20. C Beckham
    21. A Bedina Zavec
    22. A Benmoussa
    23. AC Berardi
    24. P Bergese
    25. E Bielska
    26. C Blenkiron
    27. S Bobis-Wozowicz
    28. E Boilard
    29. W Boireau
    30. A Bongiovanni
    31. FE Borràs
    32. S Bosch
    33. CM Boulanger
    34. X Breakefield
    35. AM Breglio
    36. MÁ Brennan
    37. DR Brigstock
    38. A Brisson
    39. ML Broekman
    40. JF Bromberg
    41. P Bryl-Górecka
    42. S Buch
    43. AH Buck
    44. D Burger
    45. S Busatto
    46. D Buschmann
    47. B Bussolati
    48. EI Buzás
    49. JB Byrd
    50. G Camussi
    51. DR Carter
    52. S Caruso
    53. LW Chamley
    54. YT Chang
    55. C Chen
    56. S Chen
    57. L Cheng
    58. AR Chin
    59. A Clayton
    60. SP Clerici
    61. A Cocks
    62. E Cocucci
    63. RJ Coffey
    64. A Cordeiro-da-Silva
    65. Y Couch
    66. FA Coumans
    67. B Coyle
    68. R Crescitelli
    69. MF Criado
    70. C D'Souza-Schorey
    71. S Das
    72. A Datta Chaudhuri
    73. P de Candia
    74. EF De Santana
    75. O De Wever
    76. HA Del Portillo
    77. T Demaret
    78. S Deville
    79. A Devitt
    80. B Dhondt
    81. D Di Vizio
    82. LC Dieterich
    83. V Dolo
    84. AP Dominguez Rubio
    85. M Dominici
    86. MR Dourado
    87. TA Driedonks
    88. FV Duarte
    89. HM Duncan
    90. RM Eichenberger
    91. K Ekström
    92. S El Andaloussi
    93. C Elie-Caille
    94. U Erdbrügger
    95. JM Falcón-Pérez
    96. F Fatima
    97. JE Fish
    98. M Flores-Bellver
    99. A Försönits
    100. A Frelet-Barrand
    101. F Fricke
    102. G Fuhrmann
    103. S Gabrielsson
    104. A Gámez-Valero
    105. C Gardiner
    106. K Gärtner
    107. R Gaudin
    108. YS Gho
    109. B Giebel
    110. C Gilbert
    111. M Gimona
    112. I Giusti
    113. DC Goberdhan
    114. A Görgens
    115. SM Gorski
    116. DW Greening
    117. JC Gross
    118. A Gualerzi
    119. GN Gupta
    120. D Gustafson
    121. A Handberg
    122. RA Haraszti
    123. P Harrison
    124. H Hegyesi
    125. A Hendrix
    126. AF Hill
    127. FH Hochberg
    128. KF Hoffmann
    129. B Holder
    130. H Holthofer
    131. B Hosseinkhani
    132. G Hu
    133. Y Huang
    134. V Huber
    135. S Hunt
    136. AG Ibrahim
    137. T Ikezu
    138. JM Inal
    139. M Isin
    140. A Ivanova
    141. HK Jackson
    142. S Jacobsen
    143. SM Jay
    144. M Jayachandran
    145. G Jenster
    146. L Jiang
    147. SM Johnson
    148. JC Jones
    149. A Jong
    150. T Jovanovic-Talisman
    151. S Jung
    152. R Kalluri
    153. SI Kano
    154. S Kaur
    155. Y Kawamura
    156. ET Keller
    157. D Khamari
    158. E Khomyakova
    159. A Khvorova
    160. P Kierulf
    161. KP Kim
    162. T Kislinger
    163. M Klingeborn
    164. DJ Klinke
    165. M Kornek
    166. MM Kosanović
    167. ÁF Kovács
    168. EM Krämer-Albers
    169. S Krasemann
    170. M Krause
    171. IV Kurochkin
    172. GD Kusuma
    173. S Kuypers
    174. S Laitinen
    175. SM Langevin
    176. LR Languino
    177. J Lannigan
    178. C Lässer
    179. LC Laurent
    180. G Lavieu
    181. E Lázaro-Ibáñez
    182. S Le Lay
    183. MS Lee
    184. YXF Lee
    185. DS Lemos
    186. M Lenassi
    187. A Leszczynska
    188. IT Li
    189. K Liao
    190. SF Libregts
    191. E Ligeti
    192. R Lim
    193. SK Lim
    194. A Linē
    195. K Linnemannstöns
    196. A Llorente
    197. CA Lombard
    198. MJ Lorenowicz
    199. ÁM Lörincz
    200. J Lötvall
    201. J Lovett
    202. MC Lowry
    203. X Loyer
    204. Q Lu
    205. B Lukomska
    206. TR Lunavat
    207. SL Maas
    208. H Malhi
    209. A Marcilla
    210. J Mariani
    211. J Mariscal
    212. ES Martens-Uzunova
    213. L Martin-Jaular
    214. MC Martinez
    215. VR Martins
    216. M Mathieu
    217. S Mathivanan
    218. M Maugeri
    219. LK McGinnis
    220. MJ McVey
    221. DG Meckes
    222. KL Meehan
    223. I Mertens
    224. VR Minciacchi
    225. A Möller
    226. M Møller Jørgensen
    227. A Morales-Kastresana
    228. J Morhayim
    229. F Mullier
    230. M Muraca
    231. L Musante
    232. V Mussack
    233. DC Muth
    234. KH Myburgh
    235. T Najrana
    236. M Nawaz
    237. I Nazarenko
    238. P Nejsum
    239. C Neri
    240. T Neri
    241. R Nieuwland
    242. L Nimrichter
    243. JP Nolan
    244. EN Nolte-'t Hoen
    245. N Noren Hooten
    246. L O'Driscoll
    247. T O'Grady
    248. A O'Loghlen
    249. T Ochiya
    250. M Olivier
    251. A Ortiz
    252. LA Ortiz
    253. X Osteikoetxea
    254. O Østergaard
    255. M Ostrowski
    256. J Park
    257. DM Pegtel
    258. H Peinado
    259. F Perut
    260. MW Pfaffl
    261. DG Phinney
    262. BC Pieters
    263. RC Pink
    264. DS Pisetsky
    265. E Pogge von Strandmann
    266. I Polakovicova
    267. IK Poon
    268. BH Powell
    269. I Prada
    270. L Pulliam
    271. P Quesenberry
    272. A Radeghieri
    273. RL Raffai
    274. S Raimondo
    275. J Rak
    276. MI Ramirez
    277. G Raposo
    278. MS Rayyan
    279. N Regev-Rudzki
    280. FL Ricklefs
    281. PD Robbins
    282. DD Roberts
    283. SC Rodrigues
    284. E Rohde
    285. S Rome
    286. KM Rouschop
    287. A Rughetti
    288. AE Russell
    289. P Saá
    290. S Sahoo
    291. E Salas-Huenuleo
    292. C Sánchez
    293. JA Saugstad
    294. MJ Saul
    295. RM Schiffelers
    296. R Schneider
    297. TH Schøyen
    298. A Scott
    299. E Shahaj
    300. S Sharma
    301. O Shatnyeva
    302. F Shekari
    303. GV Shelke
    304. AK Shetty
    305. K Shiba
    306. PR Siljander
    307. AM Silva
    308. A Skowronek
    309. OL Snyder
    310. RP Soares
    311. BW Sódar
    312. C Soekmadji
    313. J Sotillo
    314. PD Stahl
    315. W Stoorvogel
    316. SL Stott
    317. EF Strasser
    318. S Swift
    319. H Tahara
    320. M Tewari
    321. K Timms
    322. S Tiwari
    323. R Tixeira
    324. M Tkach
    325. WS Toh
    326. R Tomasini
    327. AC Torrecilhas
    328. JP Tosar
    329. V Toxavidis
    330. L Urbanelli
    331. P Vader
    332. BW van Balkom
    333. SG van der Grein
    334. J Van Deun
    335. MJ van Herwijnen
    336. K Van Keuren-Jensen
    337. G van Niel
    338. ME van Royen
    339. AJ van Wijnen
    340. MH Vasconcelos
    341. IJ Vechetti
    342. TD Veit
    343. LJ Vella
    344. É Velot
    345. FJ Verweij
    346. B Vestad
    347. JL Viñas
    348. T Visnovitz
    349. KV Vukman
    350. J Wahlgren
    351. DC Watson
    352. MH Wauben
    353. A Weaver
    354. JP Webber
    355. V Weber
    356. AM Wehman
    357. DJ Weiss
    358. JA Welsh
    359. S Wendt
    360. AM Wheelock
    361. Z Wiener
    362. L Witte
    363. J Wolfram
    364. A Xagorari
    365. P Xander
    366. J Xu
    367. X Yan
    368. M Yáñez-Mó
    369. H Yin
    370. Y Yuana
    371. V Zappulli
    372. J Zarubova
    373. V Žėkas
    374. JY Zhang
    375. Z Zhao
    376. L Zheng
    377. AR Zheutlin
    378. AM Zickler
    379. P Zimmermann
    380. AM Zivkovic
    381. D Zocco
    382. EK Zuba-Surma
    (2018)
    Journal of Extracellular Vesicles 7:1535750.
    https://doi.org/10.1080/20013078.2018.1535750
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52

Decision letter

  1. Richard M White
    Reviewing Editor; Memorial Sloan Kettering Cancer Center, United States
  2. Jeffrey Settleman
    Senior Editor; Calico Life Sciences, United States

In the interests of transparency, eLife includes the editorial decision letter, peer reviews, and accompanying author responses.

[Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that major issues remain unresolved.]

Decision letter after peer review:

Thank you for submitting your article "Cancer cell exosomes can initiate malignant cell transformation" 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 Jeffrey Settleman as the Senior Editor, with additional comment by Randy Schekman (Editor-in-Chief). The reviewers have opted to remain anonymous.

The Reviewing Editor has highlighted the concerns that require revision and/or responses, and we have included the separate reviews below for your consideration. If you have any questions, please do not hesitate to contact us.

Summary:

In this manuscript, the authors provide evidence that exosomes derived from pancreatic cancer cells can act as initiators of tumorigenesis. The primary evidence for this comes from a transformation assay in NIH 3T3 cells. This work adds to the growing body of evidence that exosomes can make major contributions to overall tumorigenic phenotypes.

Major concerns:

There were several major and significant concerns raised by the reviewers. They are outlined in detail below, but in general there were serious reservations about the informatic analysis and how it was performed, and a concern as to whether the mismatch repair interpretation is robust. More importantly, even if the informatics is corrected, there is no mechanistic insight into how these defects in mismatch repair occur, or even the necessity/sufficiency of this pathway.

After the review was completed, the Editor-in-Chief expressed additional concerns about the conclusions as drawn on the basis of experiments with a crude fraction of particulate material from conditioned medium:

Your claims are based on the use of "exosomes" as defined by what you refer to as established protocols for their isolation. Unfortunately, the standards in the field are inadequate and fail to account for many decades of experience in the fractionation of membrane organelles (please see Shurtleff et al., Annual Reviews of Cancer Biology, 2018). The crude particulate fraction obtained by filtration and differential centrifugation of conditioned medium contains other material including contaminating membranes and nonmembranous particles such as large RNPs. Thus, without further fractionation of the relevant sedimentable species, whether membranous or not, your conclusions about the roe of exosomes in the initiation of tumorigenesis will remain suspect.

As you are aware, in this eLife trial, we are committed to eventually publishing the paper if you are able to address these issues fully. However, given the extent of what the reviewers are requesting to make the manuscript suitable for eLife, you may want to consider withdrawing this work for further investigation. This is your decision, but please let us know your thoughts in either case.

Separate reviews (please respond to each point):

Reviewer #1:

The manuscript by Stefanius et al. reports that exosomes from a tumor cell line can initiate the process of malignant transformation in NIH-3T3 cells, which is a non-malignant stromal cell line. The authors have used exosomes from 4 pancreatic cancer cell lines as well as non-malignant immortalized pancreatic cells as an initiator in the classical cell transformation assay. The authors report that exosomes from malignant cells but not normal cells can act as initiators of cell transformation in the presence of different promoters. In addition, the exosome-transformed cells can form tumors in immunocompromised mice which histologically resemble tumors of mesenchymal origin. Further, the authors perform proteomic and exome profiling on the exosome transformed cells and observe clear differences in the protein profiles of exosome-transformed NIH-3T3 cells when compared to normal cells.

Overall, the data in Figure 1 represent an interesting finding that is not too distant from other similar findings in the exosome field, which have demonstrated that exosomes can be oncogenic in certain situations. However, the larger concern here is two fold. First, I am not convinced that the data and interpretation in Figures 2 and 3 are correct, and may be misleading for the field. Second, without a clear mechanism by which the exosomes do this, it is in the realm of phenomenology and therefore not immediately generalizable outside of this one system. This manuscript would need extensive reworking and many additional experiments to make it appropriate for eLife.

Essential revisions:

1) In Figure 1, the authors isolate foci from either MCA/TPA or ExC/TpA and then show that these grow in mice. However, they then compare this to parental 3T3 cells. This is not as useful a control since it is possible that the 3T3 culture contains cells capable of giving rise to tumors, and that what the exosomes did was select for that population. Thus, since the 3T3 cells do form foci (albeit at a smaller rate, as shown in Figure 1C) those should be the proper control, and would allow us to determine if exosome treatment is absolutely essential for tumor formation. In this model, the exosomes are not really initiators in the classic sense of the word, but instead just acting as a selection force.

2) It is unclear whether exosomes from normal pancreatic cells also have tumorigenic potential. In Figure 1C, you show that they give rise to foci at a similar rate to the DMSO control but these are never tested in the mice. This is important for the reasons related above, in that it is not clear whether the cancer cell exosome, per se, is required for tumor initiation rather than selecting for a pre-existing cell in the population.

3) The use of the 3T3 assay, while interesting and informative, does not tell us much about how general this phenomenon is. It is essential that the authors utilize other cells to tell us how unique this situation is to the 3T3. For example, they could use normal pancreas cells as well as other normal cell lines such as additional fibroblasts, mesenchymal stem cells, etc.

4) The proteomic analysis of the foci is not especially informative. The authors state that the wide variation across the samples is due to "random molecular changes across the population of treated NIH/3T3 cells". However, there is no evidence that this is really the case, since widespread variation across the samples could be due to technical artifacts from the proteomic analysis itself. How do you control for this possibility in this analysis? Moreover, the authors state "As expected the NIH/3T3 cells left untreated show no marked global changes in protein content…" yet then state that there were significant changes in the foci analysis. But how do you define what "marked" is in one context and not another in a way that is statistically rigorous? In its current form, I am not convinced this analysis is correct.

In addition, an important control experiment would be to perform the proteomic analysis on the tumors that emerged from the mice instead, since conceivably these might have more of a convergent profile that allows them to grow in the animal. In either case, it is difficult to understand how you can definitively state this is due to random effects of the exosomes without much larger numbers of samples and/or additional proteomic controls to eliminate technical artifacts.

5) In Figure 3B, the number of identified variants (more than 12,000) in each of those samples seems extraordinarily high. Were this rate of genomic alteration seen in the rest of the genome, assuming it is random mutation as the authors postulate, that would indicate that genome wide there would be nearly 600,000 mutations (i.e. if 12,000 mutations are seen in the 2% of the genome that is exonic, then one would expect 50X this number genome wide). This would be extraordinary even for highly mutated cancer such as melanoma. My expectation is that with proper stringency measures and false discovery correction, along with greater number of samples, this number of mutations would be markedly reduced.

Moreover, the authors provide no details about their use of the PLINK algorithm to cluster mutation profiles, calling into question how robust this segregation really is. This is especially concerning given the extremely high mutation burden they are seeing, which can make the interpretation of that data very problematic.

6) The interpretation that the exonic changes they are seeing is due to mismatch repair, as outlined in Figure 3 and Table 1, is not convincing. With a mutation rate this high, as outlined before, by random chance the likelihood of seeing a mutation in a mismatch repair gene is extremely high. In no way does this suggest that the mutation profile they are seeing is due to mismatch repair – the correlation is just not strong enough. The authors would absolutely have to restore mismatch repair proteins (i.e. overexpress it in the exosome treated cells) and demonstrate that this eliminates: 1) the foci formation, 2) mouse tumorigenicity, and 3) sequencing results. Otherwise, without this data, the assumption that exosomes induce random mutations via defects in mismatch repair may not be correct, and could lead the field astray.

7) Finally, the authors posit that some factor from the exosomes is inducing this initiator effect, presumably via changes in mismatch repair or possibly other mechanisms. However, no attempt to identify the mechanism for this effect is made – is it protein, mRNA, RNA, lipid, other? Although I recognize that the identification of the exosome component mediating this effect is a major challenge, without this information it is difficult to really determine the overall importance of these effects. This is an essential experiment, and likely to have major impacts on the interpretation of how exosomes act to induce oncogenic changes in nearby cells.

Additional data files and statistical comments:

The proteomic and exonic analysis is not convincing in its current form.

Reviewer #2:

The manuscript by Stefanius and colleagues provides further evidence that cancer cell derived exosomes can initiate events that can lead to the transformation of NIH3T3 cells. Exosomes from cancer cells irrespective of the Kras mutation status can act as initiators that prime the promotors to transform cells and enable tumor formation by the NIH3T3 cells. This is an interesting paper but it is unclear as to its novelty and innovation as presented. The study has potential. Some comments are provided below.

1) While the use of MCA and TPA has been around for a while, it is not clear what the in vivo relevance is at this point. It would more relevant to use nonchemical promoters not generally found in human tissue. This should be attempted. Would the exosomes initiate NIH3T3 cells if they are injected into mice with their pancreas primed using relevant genetic promoters, and induce tumor formation? Such experiments would move the field forward. The basic discovery that exosomes can induce transformation of cells without such promoting agents (TPA) has been published and should be cited. Some studies actually use serum exosomes from cancer patients to illustrate their ability to induce tumor formation by cells.

http://www.pnas.org/content/108/12/4852/

https://www.ncbi.nlm.nih.gov/pubmed/25446899/

https://www.ncbi.nlm.nih.gov/pubmed/28854931/

https://www.ncbi.nlm.nih.gov/pubmed/27179759/

2) It is unclear why foci were used to induce tumors versus the cells by themselves after the exosomes mediated initiation and TPA induced promotion. This should be addressed.

3) The amount of exosomes used is not clear and whether it represents physiologically relevant concentrations. What does 80ng/mL represent in particle number? Also, it is critical to perform dose response experiments in this setting.

4) The exclusive use of NIH3T3 limits the implication of the study. These are transformed cells. This study could include use of other cells of different lineages and just fibroblasts. The use of primary cells would make it relevant to human cancers.

5) More control lines others than just HPDE would make the observations more interesting.

6) Mechanistic insights are minimal in this study considering what has already been published. Important questions remain such as are exosomes initiating the purported events, and why is the identification of the mediators important. Why is the DNA damage and mismatched events induced, and why are these similar to MCA need a more experimental explanation. Gain of function and loss of function experiments are important.

7) The fact that the tumors did not identify any changes is interesting and needs to be further addressed.

Reviewer #3:

This is a potentially interesting study that clearly shows that (1) Cancer-derived exosomes effectively initiate malignant transformation in mouse NIH3T3 cells in vitro; (2) The transformed NIH3T3 cells (capable of foci formation) form subQ fibrosarcomas; and (3) not unexpectedly, exosome-induced tumor initiation is related to defective DNA repair, however, the repair defects are random (not caused by one specific exosome-associated oncogene).

The manuscript is well executed, and provides multiple controls, such as additional promoters, transposition of initiator and promoter in two-step carcinogenesis assay, etcetera. However, the study its limited to investigation of NH3T3 transformation and these cells are notoriously easy to transform. Also, the biological relevance of the study is somewhat questionable and critical issues remain unresolved.

First, could a similar process occur in vivo, in mice or humans? If so, an individual with a primary pancreatic tumor should have a higher propensity to form other primary tumor(s). Is that the case? Do such individuals have a chance of exposure to promoters? Second, could this phenomenon be extended to other tumor types (not PDAC)? And last, would exosome-initiated tumors be exclusively fibrosarcomas? Other tumor types? Random tumors? Would PDAC exosomes act as tumor initiating agent in pancreatic epithelium?

These questions could be answered only by using PDAC exosomes in a classical two-step carcinogenesis assay, as opposed to in vitro transformation assay followed by tumorigenesis assay. Similar experiments demonstrating transformation events due to the transfer of exosomes' contents to the bystander cells have already been performed using exposures isolated from the tumor cell lines or from the sera of cancer patients (please see reference below). Most of these experiment are performed as a one step process, which also contradicts the conclusions from the present analysis.

Finally, the potential ability of exosomes to promote tumor growth by virtue of transforming host cells could be tested using fluorescent mice (GFP, RFP etc) as donors or acceptors of exosome-producing NIH-3T3 cells. The appearance of fluorescence tagged cells in a non-tagged exotic tumor formed by exosome-transformed NIH-3T3 cells, whose origin could be ascertained using other fluorescent makers would clearly ascertain such recruitment.

Regrettably, this potentially interesting study is insufficiently novel to be published in eLife in its present form, given limited scope of analyses performed and potential low relevance of the results.

In favor of the potential relevance of this study, 18% of co-occurring IPMN and ductal adenocarcinomas were likely independent, suggesting that the carcinoma arose from an independent precursor. By contrast, all colloid carcinomas were likely related to their associated IPMNs. However, to link this process to exosome-dependent tumor initiation, one has to analyze exosomes from patients from IPMNs and their potential capacity to induce pancreatic tumors. (Felsenstein et al). However, the present study is extremely preliminary in nature and does not unequivocally address such a possibility. Further experiments are needed to demonstrate the tumor-initiating capacity of exosomes from the neoplastic foci.

https://www.nature.com/articles/s41419-018-0485-1.pdf (Daie et all, Cell Death and Disease (2018)9:454); https://jeccr.biomedcentral.com/track/pdf/10.1186/s13046-017-0587-0 (Abdouh et al., Journal of Experimental & Clinical Cancer Research (2017) 36:113);

http://www.pnas.org/content/108/12/4852 (Antonyak et al., 2011. PNAS 108 (12) 4852-4857); https://doi.org/10.1016/j.bbrc.2014.07.109 (Lee et al., BBRC Volume 451, Issue 2, 22 August 2014, Pages 295-301

http://cancerres.aacrjournals.org/content/77/21/5808.full-text.pdf (Figueroa et al. Cancer Res. 2017 Nov 1;77(21):5808-5819. doi

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4254633/pdf/nihms631122.pdf (Melo et al., Cancer Cell. 2014 November 10; 26(5): 707-721)

Additional data files and statistical comments:

The experiments as presented are performed rigorously, with appropriate statistical analysis, however the conclusions that could be drawn are limited and insufficiently novel.

Evaluation of the paper after the first round of revision:

Thank you for submitting your article "Human pancreatic cancer cell exosomes, but not human normal cell exosomes, act as an initiator in cell transformation" for consideration by eLife.

While you have performed a few additional experiments that strengthen the manuscript, the reviewers agree that many of the original criticisms remain unaddressed. As you are aware, in this eLife trial we are committed to ultimately publishing the paper should you choose to proceed with it. However, all of the reviewer and editorial concerns about the validity of these findings will be published in parallel with the paper.

There remain several major areas of concern:

1) The methods used to isolate the exosomes.

Randy Schekman specifically commented upon the new data as follows:

"I have examined the additional data relevant to the concern I raised about the crude fraction they use to claim an exosome activity. As far as I can tell this consists of one data point with a peak fraction of buoyant membranes assayed at one concentration. This would not pass muster by any standard biochemical analysis. I recommend assays across the peak fraction of exosomal marker proteins and a quantitative titration of activity vs. protein concentration in the peak fraction in comparison to the crude starting material. The expectation is that if this activity tracks with exosomes, the specific activity in their biological assay should have been enriched by the buoyant density separation. Without this it is premature to conclude that the activity is associated with exosomes."

2) The notion that exosomes are initiators.

Several of the reviewers raised concerns about using 3T3 cells as the sole evidence of initiation activity. This was not addressed experimentally, but instead the authors state that "It will be very interesting to test this function in other cell types, and will be done in future studies." Despite this being a major concern, this is still left unaddressed.

Why this is important is related to the new data in Figure 3B, Figure 3—figure supplement 2A, they show that the background foci actually do form tumors, albeit at a slower rate and with lower efficiency than the exosome treated 3T3 foci. But the fact that the background foci do indeed form tumors actually argues against a pure initiator activity – it seems to accelerate what is already happening in the background foci anyway. Thus, a central argument that these exosomes are true initiators does not seem fully accurate. In the absence of other cell types other than 3T3s, it is hard to state this definitively outside of this one context.

3) The proteomic analysis.

While the authors make it clear how they performed the analysis, it is still confusing to state "However, proteins found in transformed cells shown in Figure 4 exhibit more divergence from control cells" Divergence based on what statistical measure? Venn diagrams are not statistical methods, and it is not clear that the PANTHER data shows true statistical enrichment in the exosome treated cells.

4) Exome analysis.

In the rebuttal, the authors state that "we feel confident that the method we used is sufficient to obtain reliable results". Yet the request to analyze this data with additional pipelines and more stringent statistics was not performed. There is no orthogonal validation of this pipeline to ensure that the results are accurate.

In looking at their more detailed methods, it is now clear that they are comparing their exomes to the human reference genomes (GRCm38), so that the variants being called are actually just variants that differ from this reference genome, and not the parental 3T3 cells. How do you know that most of the called variants are not just called SNPs or baseline point mutations between 3T3 cells and the human reference genome? Why not compare the exome data to the parental 3T3 cells rather than the human reference genome?

The authors now provide additional data, using the same pipelines, on bulk exome data on both untreated and TPA treated foci in Figures 5 and 6. In Figure 5—figure supplement 1, they essentially show that the number of mutations in the untreated or TPA treated cells is similar to that in the MCA/TPA or ExC/TPA foci. But if the basic idea is that exosomes act as initiators through increases in mutation rate, then why are the number of mutations essentially the same. Furthermore, in the data in Figure 6, you show that the signature 20 is higher in the MCA/TPA or ExC/TPA samples compared to untreated/TPA, but this just looks like a modest increase, and there is no statistical analysis to say whether these are really different. How do you confirm that you are really enriching for this signature in this dataset other than a heatmap which looks to be so.

Finally, it is not clear that bulk sequencing of this population of highly mutated cells is appropriate. If we consider that each cell could have many different mutations, the more appropriate experiment is to create single cell clones out of the transformed foci, let that grow up, and then perform the exome analysis on these isolated clones. While this is an expensive and ultimately very time consuming method, it would ultimately be much more informative in understanding how the mutations occur in each cell, rather than a population of cells which may or may not have defects is mismatch repair.

5) The role of mismatch repair.

While we recognize that adding in single mismatch repair genes may not rescue the phenotypes, at the same time the authors are making a very bold claim that exosomes specifically lead to MMR defects. This is perhaps the most intriguing idea in the paper, but without any experiments to mechanistically dissect this it remains correlative.

Evaluation of the paper after the second round of revision:

In this revised manuscript, the authors have tried to address several of the issues brought up during the review process. As part of this trial, we will proceed with publication of the manuscript although we do feel several issues remain unresolved that could have further strengthened the paper.

First, Dr Schekman raised a point that dose response relationships comparing the more purified exosomes with the crude starting material would be required to establish that the buoyant density gradient achieved purification of the active vesicle species. The single concentration of the gradient fraction used in the new version of the paper fails to establish meaningful purification.

Second, while we agree that the 3T3 assay is a standard initiator assay, the development of other assays to confirm this potentially new and important function of exosomes would have provided significant confidence that this is a generalizable activity.

Third, the authors have used one informatic pipeline for their analysis, whereas adding in additional and complementary pipelines for mutation calling (i.e MuTect, etc) would have provided increased confidence that these are reproducible results from an informatic standpoint. Analysis of single cell clones from the transformed cells would have also given a more fine-grained analysis of the mutational pattern induced via exosomes, which cannot be achieved by bulk sequencing alone.

https://doi.org/10.7554/eLife.40226.031

Author response

Major concerns:

There were several major and significant concerns raised by the reviewers. They are outlined in detail below, but in general there were serious reservations about the informatic analysis and how it was performed, and a concern as to whether the mismatch repair interpretation is robust. More importantly, even if the informatics is corrected, there is no mechanistic insight into how these defects in mismatch repair occur, or even the necessity/sufficiency of this pathway.

Below we have addressed point by point the concerns of each reviewer. Text has been added to explain the informatic analysis to assure the reviewers that there is rigor in all methods used for our conclusions. Additionally, the informatics analysis, particularly the exome sequencing data, has been expanded upon as described in detail below. As mentioned in the title, summation, and manuscript, we have identified, for the first time, that cancer cell exosomes can act as classic initiators in the cell transformation. Our studies provide molecular mechanistic insight by demonstrating that cancer cell exosomes, like other classic chemical initiators, induce random mutations into cells. Cells from this pool of randomly mutated cells can then be driven to transformation by a classic promoter.

After the review was completed, the Editor-in-Chief expressed additional concerns about the conclusions as drawn on the basis of experiments with a crude fraction of particulate material from conditioned medium:

Your claims are based on the use of "exosomes" as defined by what you refer to as established protocols for their isolation. Unfortunately, the standards in the field are inadequate and fail to account for many decades of experience in the fractionation of membrane organelles (please see Shurtleff et al., Annual Reviews of Cancer Biology, 2018). The crude particulate fraction obtained by filtration and differential centrifugation of conditioned medium contains other material including contaminating membranes and nonmembranous particles such as large RNPs. Thus, without further fractionation of the relevant sedimentable species, whether membranous or not, your conclusions about the roe of exosomes in the initiation of tumorigenesis will remain suspect.

We agree that one of the current challenges in the exosome field is the lack of a universally accepted purification method. Discussion surrounding this and the current standards used to validate exosome isolation is included in the text. Exosomes were characterized according to guidelines laid out in the Minimal Information for Studies of Extracellular Vesicles (MISEV2018). Additionally, in Figure 2—figure supplement 1B, we show that further fractionation by sucrose density gradient does not change the transformation potential of cancer cell exosomes.

Discussion section:

“To eliminate the possibility that contaminants might participate in the exosome initiator activity, we further purified the exosomes using a sucrose density gradient in order to obtain ‘pure’ exosome fractions (32). Identical experiments with both populations of exosomes demonstrated that both samples could act as an initiator in the CTA, thereby establishing exosomes as the initiating factor in this assay (Figure 2—figure supplement 1B).”

Separate reviews (please respond to each point):

Reviewer #1:

[…] Overall, the data in Figure 1 represent an interesting finding that is not too distant from other similar findings in the exosome field, which have demonstrated that exosomes can be oncogenic in certain situations. However, the larger concern here is two fold. First, I am not convinced that the data and interpretation in Figures 2 and 3 are correct, and may be misleading for the field. Second, without a clear mechanism by which the exosomes do this, it is in the realm of phenomenology and therefore not immediately generalizable outside of this one system. This manuscript would need extensive reworking and many additional experiments to make it appropriate for eLife.

Essential revisions:

1) In Figure 1, the authors isolate foci from either MCA/TPA or ExC/TpA and then show that these grow in mice. However, they then compare this to parental 3T3 cells. This is not as useful a control since it is possible that the 3T3 culture contains cells capable of giving rise to tumors, and that what the exosomes did was select for that population. Thus, since the 3T3 cells do form foci (albeit at a smaller rate, as shown in Figure 1C) those should be the proper control, and would allow us to determine if exosome treatment is absolutely essential for tumor formation. In this model, the exosomes are not really initiators in the classic sense of the word, but instead just acting as a selection force.

To address this concern, we performed in vivo experiments using background foci from untreated NIH/3T3 cells. Three independent background foci from different wells were isolated, expanded, and injected into NSG (NOD scid γ) mice (n=5).

Subsection “In vivo studies confirm the fully transformed state of cancer cell exosome-initiated cells”:

“After injection of untreated background foci, we observed that six out of 15 total mice formed tumors. Notably, each of the six tumors (5 from the same injected foci) grew later in the time course and at a significantly slower rate compared to initiator-promoter treated transformed foci cells (Figure 3B, Figure 3—figure supplement 2A).”

2) It is unclear whether exosomes from normal pancreatic cells also have tumorigenic potential. In Figure 1C, you show that they give rise to foci at a similar rate to the DMSO control but these are never tested in the mice. This is important for the reasons related above, in that it is not clear whether the cancer cell exosome, per se, is required for tumor initiation rather than selecting for a pre-existing cell in the population.

Quantification of all cell transformation assays shown in Figure 2 and Figure 2—figure supplement 1 consistently show that a uniform level of background foci are formed in all control experiments. This is likely due to the nature of NIH/3T3 cells, as it has been shown that when NIH/3T3 cells are maintained as confluent, non-growing cultures without passage, or injected into mice in high concentrations, they may undergo spontaneous cell transformation (1). Therefore, when stringent focus scoring is carried out according to established criteria (Materials and methods), a constant level of background foci are observed in each experiment. Only in experiments where cells were treated with a functional initiator and promoter, did we observe a significant increase in transformation, resulting in formation of an additional 3 to 4 foci per well. This is why background foci from untreated cells were injected into mice as a control to assess the tumor forming potential of these background foci, as described above.

3) The use of the 3T3 assay, while interesting and informative, does not tell us much about how general this phenomenon is. It is essential that the authors utilize other cells to tell us how unique this situation is to the 3T3. For example, they could use normal pancreas cells as well as other normal cell lines such as additional fibroblasts, mesenchymal stem cells, etc.

Explanation and reasoning behind our use of the classic 2-stage cell transformation assay was added to the Introduction section:

“Malignant transformation of a normal cell occurs in a stepwise fashion. Point mutations in the genome can result in the reprogramming of a normal cell to a less differentiated state that is receptive to additional genetic alterations resulting in uncontrolled growth and ultimately cancer. The classic 2-stage in vitro cell transformation assay (CTA) is a tiered system for transformation that was created for screening potential carcinogenic factors (21-23). In this system, cells are first treated with a suspected carcinogen, called an initiator, such as the genotoxic carcinogen 3-MCA (3-methylcholanthrene). 3-MCA introduces random genetic changes in a pool of normal cells. Subsequently, these initiated cells are exposed to a promoter, such as TPA (12-O-tetradecanoylphorbol 13-acetate), which enhances proliferation in the initiated cells selectively, thus driving malignant transformation of the cells. The resulting transformed cells are observed as foci on a cell culture plate (23, 24). This reductionist approach provides sensitivity in detecting a wider range of initiating agents that may not show obvious transforming activity without a promoter (23). Using this assay as a model system for malignant cell transformation, we assessed whether or not cancer cell-derived exosomes could affect and/or potentially drive the transformation of a normal cell. The results presented herein provide a detailed analysis of a previously unidentified molecular function of cancer cell exosomes for malignant cell transformation.”

It will be very interesting to test this function in other cell types, and will be done in future studies. As the optimization for another assay (2-3 months/assay condition) to be equally reliable as the CTA with 3T3-cells will require over a year, we think this experiment is beyond the scope of this paper.

4) The proteomic analysis of the foci is not especially informative. The authors state that the wide variation across the samples is due to "random molecular changes across the population of treated NIH/3T3 cells". However, there is no evidence that this is really the case, since widespread variation across the samples could be due to technical artifacts from the proteomic analysis itself. How do you control for this possibility in this analysis? Moreover, the authors state "As expected the NIH/3T3 cells left untreated show no marked global changes in protein content…" yet then state that there were significant changes in the foci analysis. But how do you define what "marked" is in one context and not another in a way that is statistically rigorous? In its current form, I am not convinced this analysis is correct.

Specific details pertaining to how the proteomic analysis (Figure 4 and Figure 4—figure supplement 1) are discussed in detail in Materials and methods. Samples were analyzed using a common LC-MS/MS method and appropriate steps were taken to ensure accuracy of results and avoid potential “technical artifacts,” including proper sample prep, instrument calibration, n=3 replicates, PSM filtering <0.01, protein filtering <0.01, etc. The assessment that “no marked global changes in protein content were observed for [initiated] cells treated with MCA or Capan-2 exosomes when compared to proteins found in untreated NIH/3T3 cells” is based on analysis showing that 80% of the total proteins found (2287/2859) were identified in all three conditions, while only 9.7% of the total proteins found (276/2859) were identified in only a single condition (Figure 4—figure supplement 1). However, proteins found in transformed cells shown in Figure 4 exhibit more divergence from control cells. We have altered the presentation of data in Figure 4 to highlight that the differences in protein content between foci result primarily from the contributions of “unique proteins” that are not found in the control cells. To probe these differences further, we performed Gene Ontology enrichment analysis on the proteins found exclusively in transformed foci (Figure 4D).

In addition, an important control experiment would be to perform the proteomic analysis on the tumors that emerged from the mice instead, since conceivably these might have more of a convergent profile that allows them to grow in the animal. In either case, it is difficult to understand how you can definitively state this is due to random effects of the exosomes without much larger numbers of samples and/or additional proteomic controls to eliminate technical artifacts.

We agree that performing proteomic analysis of the tumors that emerged from mice would be an interesting experiment, however we believe it is beyond the scope of this paper. We were primarily interested in determining if any changes could be detected at the protein level as a direct result of either treatment with an initiator for three days or formation of foci from cell transformation. Since we were working in the context of NIH/3T3 cells, each treatment condition could then be compared to background proteins present in untreated NIH/3T3 cells as a control. Additionally, we have changed the language in the text to clarify that we cannot directly attribute changes observed in protein content of transformed foci to random effects of the exosome treatment. We can only compare the protein profiles of NIH/3T3 cells directly after initiation treatment to those of cells after they have transformed.

5) In Figure 3B, the number of identified variants (more than 12,000) in each of those samples seems extraordinarily high. Were this rate of genomic alteration seen in the rest of the genome, assuming it is random mutation as the authors postulate, that would indicate that genome wide there would be nearly 600,000 mutations (i.e. if 12,000 mutations are seen in the 2% of the genome that is exonic, then one would expect 50X this number genome wide). This would be extraordinary even for highly mutated cancer such as melanoma. My expectation is that with proper stringency measures and false discovery correction, along with greater number of samples, this number of mutations would be markedly reduced.

Moreover, the authors provide no details about their use of the PLINK algorithm to cluster mutation profiles, calling into question how robust this segregation really is. This is especially concerning given the extremely high mutation burden they are seeing, which can make the interpretation of that data very problematic.

First, we expanded the Materials and methods section 2.6 on exome sequencing analysis to provide more detailed description about our strategy to analyze data obtained during exome sequencing and how all bioinformatics was performed by collaborators and bioinformatic experts, Min S. Kim and Venkat Malladi. We do acknowledge that more stringent criteria (read depth of DP higher than 10) and use of additional programs together with Strelka2 and combining results to obtain consensus from different variant callers could reduce the number of identified variants, but we feel confident that the method we used is sufficient to obtain reliable results.

Second, we do agree with the reviewer’s math, when it pertains to cells not carrying any mutations in genes that would lead to genetic instability at the nucleotide level. In this situation, each round of cell division would not result in an extreme number of new mutations. Based on COSMIC signatures we know that mismatch repair dysfunction is contributing to mutational burden of transformed cells from MCA/TPA and Capan-2/TPA foci leading to possibility where every cell division can amplify the number of mutations exponentially. When analyzing mismatch repair genes in more details, missense mutations were found to be encoded in each of the transformed foci supporting the finding of dysfunctional MMR system. We also see this signature in background foci, although not the top signature which in these cases was COSMIC Signature 3. This signature is associated with failure of DNA double-strand break-repair and would also contribute to a high mutation rate supporting the high number of variants found in all samples.

6) The interpretation that the exonic changes they are seeing is due to mismatch repair, as outlined in Figure 3 and Table 1, is not convincing. With a mutation rate this high, as outlined before, by random chance the likelihood of seeing a mutation in a mismatch repair gene is extremely high. In no way does this suggest that the mutation profile they are seeing is due to mismatch repair – the correlation is just not strong enough. The authors would absolutely have to restore mismatch repair proteins (i.e. overexpress it in the exosome treated cells) and demonstrate that this eliminates: 1) the foci formation, 2) mouse tumorigenicity, and 3) sequencing results. Otherwise, without this data, the assumption that exosomes induce random mutations via defects in mismatch repair may not be correct, and could lead the field astray.

To address this concern we performed additional exome sequencing analysis for three untreated background foci and three TPA- only treated background foci. Figure 6 highlights mutational signatures found in all 12 samples. These additional sequencing results show that COSMIC signature associated with defective DNA mismatch repair, is not the top signature found in the background foci and it gives us more confidence that our result is reliable. In addition, analysis of the mismatch repair genes showed that no mutations were found in three background foci and only one TPA-only focus had mutations, supporting the probability of MMR dysfunction in the transformed foci. We also provide a more detailed explanation on how all bioinformatics was performed in Materials and methods.

Moreover, the interpretation of a COSMIC profile for mismatch repair and genetic instability is valid based on the published methods for analysis of tumor profiles using exosome sequencing.

Ref 37. A. Goncearenco et al., Exploring background mutational processes to decipher cancer genetic heterogeneity. Nucleic Acids Res 45, W514-w522 (2017).

The experiment suggested by the reviewer to rescue the transformation with wild type copies of mutated genes is not feasible. The type and number of mutations in any one cell caused by mismatch repair and microsatellite instability are large in number and vary from one cell to another. Even cloning cells from a population leads to variation (Orth, K., et al. PNAS USA (1994) 91(20), 9495-9.)

These cells, as they are transformed with multiple mutations, cannot be corrected with the addition of just one gene.

7) Finally, the authors posit that some factor from the exosomes is inducing this initiator effect, presumably via changes in mismatch repair or possibly other mechanisms. However, no attempt to identify the mechanism for this effect is made – is it protein, mRNA, RNA, lipid, other? Although I recognize that the identification of the exosome component mediating this effect is a major challenge, without this information it is difficult to really determine the overall importance of these effects. This is an essential experiment, and likely to have major impacts on the interpretation of how exosomes act to induce oncogenic changes in nearby cells.

The experiments suggested by the reviewer are of interest to us, but would be a major challenge, as stated. It is known that exosomes contain heterogeneous cargo and that no two exosomes contain the same set of (see below). More work on characterizing the exosomes is necessary and we believe this is out of the scope of this specific study.

We have added text to address this issue in the Discussion section:

“A recent study by Felsentein et al. described an interesting finding in which a portion of co-occurring IPMNs in PDAC patients appeared to be genetically unrelated, meaning they shared no mutations in the assayed genes (42). This elicits a fascinating question as to whether cells in a primary tumor could be derived from independent transformation events as opposed to exclusive clonal events (42). Potentially, exosomes secreted by the primary tumor could orchestrate such events. Additional studies towards understanding the cancer cell exosome mediated initiation are needed to address such questions, specifically, investigating the initiation capacity of cancer cell exosomes in more relevant cells like human epithelial cells.”

And:

“Additionally, it is known that exosomes do not uniformly contain the same pool of proteins, lipids, metabolites, or microRNAs, but rather each exosome contains a unique repertoire of biological molecules thus possibly causing a variety of changes in the cells (7, 9, 12). Therefore, potentially both MCA and cancer cell exosomes are causing random molecular changes across the population of treated NIH/3T3 cells that cannot be detected amongst the overall protein composition of cells by mass spectrometry. Ultimately, more studies are needed to elucidate the molecular mechanism of this cancer cell exosome-mediated initiation event.”

Additional data files and statistical comments:

The proteomic and exonic analysis is not convincing in its current form.

We have addressed this concern in our Materials and methods and demonstrate rigor in our analyses. Please see above comments specifically regarding proteomic and exome sequencing data analysis.

Reviewer #2:

The manuscript by Stefanius and colleagues provides further evidence that cancer cell derived exosomes can initiate events that can lead to the transformation of NIH3T3 cells. Exosomes from cancer cells irrespective of the Kras mutation status can act as initiators that prime the promotors to transform cells and enable tumor formation by the NIH3T3 cells. This is an interesting paper but it is unclear as to its novelty and innovation as presented. The study has potential. Some comments are provided below.

1) While the use of MCA and TPA has been around for a while, it is not clear what the in vivo relevance is at this point. It would more relevant to use nonchemical promoters not generally found in human tissue. This should be attempted. Would the exosomes initiate NIH3T3 cells if they are injected into mice with their pancreas primed using relevant genetic promoters, and induce tumor formation? Such experiments would move the field forward. The basic discovery that exosomes can induce transformation of cells without such promoting agents (TPA) has been published and should be cited. Some studies actually use serum exosomes from cancer patients to illustrate their ability to induce tumor formation by cells.

http://www.pnas.org/content/108/12/4852 https://www.ncbi.nlm.nih.gov/pubmed/25446899/

https://www.ncbi.nlm.nih.gov/pubmed/28854931/

https://www.ncbi.nlm.nih.gov/pubmed/27179759/

Explanation and reasoning behind our use of the classic 2-stage cell transformation assay was added to the Introduction section:

“Malignant transformation of a normal cell occurs in a stepwise fashion. Point mutations in the genome can result in the reprogramming of a normal cell to a less differentiated state that is receptive to additional genetic alterations resulting in uncontrolled growth and ultimately cancer. The classic 2-stage in vitro cell transformation assay (CTA) is a tiered system for transformation that was created for screening potential carcinogenic factors (21-23). In this system, cells are first treated with a suspected carcinogen, called an initiator, such as the genotoxic carcinogen 3-MCA (3-methylcholanthrene). 3-MCA introduces random genetic changes in a pool of normal cells. Subsequently, these initiated cells are exposed to a promoter, such as TPA (12-O-tetradecanoylphorbol 13-acetate), which enhances proliferation in the initiated cells selectively, thus driving malignant transformation of the cells. The resulting transformed cells are observed as foci on a cell culture plate (23, 24). This reductionist approach provides sensitivity in detecting a wider range of initiating agents that may not show obvious transforming activity without a promoter (23). Using this assay as a model system for malignant cell transformation, we assessed whether or not cancer cell-derived exosomes could affect and/or potentially drive the transformation of a normal cell. The results presented herein provide a detailed analysis of a previously unidentified molecular function of cancer cell exosomes for malignant cell transformation.”

It will be very interesting to test this function in other cell types, and will be done in future studies. As the optimization for another assay (2-3 months/assay condition) to be equally reliable as the CTA with 3T3-cells will require over a year, we think this experiment is beyond the scope of this paper.

We have cited the suggested papers and incorporated discussion on the role that exosomes have been previously shown to play in cell transformation in the Discussion section:

“Recent studies demonstrate that cancer cell exosomes can in fact participate in cell transformation, however the details on how these exosomes contribute to transformation remained elusive. For example, Dai et al. showed that exosomes participated in the transformation process of normal cells after cells were treated with arsenite (15). Another study described a transfer of malignant traits to BRCA1 deficient human fibroblast when treated with either cancer patient sera or isolated cancer cell exosomes, leading to malignant transformation (16, 17). In addition, exosomes in breast cancer patient sera have been shown to promote nontumorigenic epithelial cells to form tumors in a Dicer-dependent manner (11). In each of the aforementioned studies, non-transformed cells were either pre-exposed to transforming agents or treated with cancer patient sera or medium from cultured cancer cells. Interestingly, Antonyak et al. has demonstrated that sustained treatment of NIH/3T3 cells with breast cancer or glioma cell-derived microvesicles has the ability to induce transformation of recipient cells (18). Contrastingly, in our studies, when cells were treated only with pancreatic cancer cell exosomes, increased cell transformation was not observed. While the specific details and effects on cell transformation attributed to exosomes vary between the studies, both demonstrate that exosomes are indeed playing a role in cell transformation.”

2) It is unclear why foci were used to induce tumors versus the cells by themselves after the exosomes mediated initiation and TPA induced promotion. This should be addressed.

In the 2-stage cell transformation assay, the phenotype for transformation is focus formation, specifically representing the transformed cells among the total cell population. For this reason, foci were used for in vivo assays. This specific issue is added to the Discussion section:

“The classic 2-stage CTA was specifically used as a tool to study cell transformation because it is rigorous, reproducible, and has a well-established phenotype.”

Additionally, we wanted to assess the heterogeneity between foci formed from the same treatment. This is why multiple independent foci of the same type were compared for all treatment conditions. Discussion on the use of foci cells in the in vivo assay is stated in the text in subsection “In vivo studies confirm the fully transformed state of cancer cell exosome-initiated cells”:

“An important step of assessing the tumorigenic property of transformed cells is their ability to form tumors in vivo. To determine whether exosome-initiated transformed cells have the capacity to form tumors when injected subcutaneously into immunocompromised mice, we first isolated and expanded foci cells from the MCA/TPA and Capan-2 exosome/TPA experiments (Figure 3A).”

“Additional in vivo studies were performed to analyze the tumor forming potential of a variety of foci, including background foci formed in control experiments. (Figure 3, Figure 3—figure supplement 1, Figure 3—figure supplement 2).”

3) The amount of exosomes used is not clear and whether it represents physiologically relevant concentrations. What does 80ng/mL represent in particle number? Also, it is critical to perform dose response experiments in this setting.

We agree and to address this we performed nanoparticle tracking analysis and based on this analysis, 80ng/mL is equivalent to 7.0x107 particles/mL (Figure 1, Figure 1—figure supplement 1).

In addition, we have performed dose response experiments and added results to Figure 2—figure supplement 1C. Results are discussed in the text:

“Dose-response studies were performed using protein concentration as a normalization strategy to evaluate the amount of cancer cell exosomes needed to initiate cell transformation. As a standard concentration in each transformation assay we used 80 ng/mL of proteins, corresponding to 7.0x107 particles/mL. Exosome protein concentrations ranging from 0.08 ng/mL to 2400 ng/mL were tested and we observed equal cell transformation for all concentrations with the exception of the two lowest, 0.08 ng/mL and 0.8 ng/mL. This indicates that initiator activity of cancer cell exosomes requires one dose of exosomes over a 3-day period with a protein concentration of at least 8 ng/mL (Figure 2—figure supplement 1C, Figure 2—figure supplement 1-source data 3).”

A reliable estimate of the number of particles made by cancer cells in vivo is not readily available, but based on own data we observe over a two log increase in exosome production from cancer cells compared to normal cells in vitro

4) The exclusive use of NIH3T3 limits the implication of the study. These are transformed cells. This study could include use of other cells of different lineages and just fibroblasts. The use of primary cells would make it relevant to human cancers.

Explanation and reasoning behind our use of the classic 2-stage cell transformation assay was added to the Introduction section:

“Malignant transformation of a normal cell occurs in a stepwise fashion. Point mutations in the genome can result in the reprogramming of a normal cell to a less differentiated state that is receptive to additional genetic alterations resulting in uncontrolled growth and ultimately cancer. The classic 2-stage in vitro cell transformation assay (CTA) is a tiered system for transformation that was created for screening potential carcinogenic factors (21-23). In this system, cells are first treated with a suspected carcinogen, called an initiator, such as the genotoxic carcinogen 3-MCA (3-methylcholanthrene). 3-MCA introduces random genetic changes in a pool of normal cells. Subsequently, these initiated cells are exposed to a promoter, such as TPA (12-O-tetradecanoylphorbol 13-acetate), which enhances proliferation in the initiated cells selectively, thus driving malignant transformation of the cells. The resulting transformed cells are observed as foci on a cell culture plate (23, 24). This reductionist approach provides sensitivity in detecting a wider range of initiating agents that may not show obvious transforming activity without a promoter (23). Using this assay as a model system for malignant cell transformation, we assessed whether or not cancer cell-derived exosomes could affect and/or potentially drive the transformation of a normal cell.

The results presented herein provide a detailed analysis of a previously unidentified molecular function of cancer cell exosomes for malignant cell transformation.”

It will be very interesting to test this function in other cell types, and will be done in future studies. As the optimization for another assay (2-3 months/assay condition) to be equally reliable as the CTA with 3T3-cells will require over a year, we think this experiment is beyond the scope of this paper.

5) More control lines others than just HPDE would make the observations more interesting.

As suggested by the reviewer, we have added an additional “normal cell” exosome control in addition to HPDE cells. Exosomes were isolated from human primary dermal fibroblasts and tested in the cell transformation assay as an initiator. Results show that like HPDE exosomes, exosomes from primary fibroblasts are unable to initiate cell transformation above background levels. Please refer to Figure 2—figure supplement 1A.

6) Mechanistic insights are minimal in this study considering what has already been published. Important questions remain such as are exosomes initiating the purported events, and why is the identification of the mediators important. Why is the DNA damage and mismatched events induced, and why are these similar to MCA need a more experimental explanation. Gain of function and loss of function experiments are important.

First, as mentioned in the title, summation and manuscript, we have identified, for the first time, that cancer cell exosomes can act as classic initiators in the cell transformation. We have cited previous studies and incorporated discussion on the role that exosomes have been previously shown to play in cell transformation in the Discussion. Our studies provide mechanistic insight by demonstrating that cancer cell exosomes, like other classic chemical initiators, induce random mutations into cells. Cells from this pool of randomly mutated cells can then be driven to transformation by a classic promoter.

Second, we also were surprised that MCA/TPA and Capan-2 exosome/TPA transformed foci have similar mutational signatures. To understand this better, we performed additional exome sequencing studies for three untreated background foci and three TPA- only treated background foci to learn whether the same mutational signature is seen in foci without any treatment or treatment with the TPA, as these were the two common factors (NIH-3T3 cells and TPA) with our previous samples. Our exosome analysis revealed that all of the sequenced background foci (6) had a similar mutagenic profile whereas none of the transformed foci encoded a similar mutagenic profile. These additional results show that mismatch repair is a main driver for mutations found in transformed cells and that the COSMIC signature associated with defective DNA mismatch repair is not the top signature found in the background foci. We discuss these findings in the Discussion:

“Moreover, even though all six transformed foci were derived from treatment with a common promoter (TPA), foci from TPA-only treated cells do not show the same mutational signatures as the initiator/promoter transformed foci. This supports a proposal that when TPA is used as a promoter, the molecular path towards tumorigenesis is not random, but rather TPA will drive a path using randomly mutated cells that accommodates a course towards a Cosmic 20 or 15 profile. Furthermore, future studies might reveal that another promoter acting on initiator treated cells might lead towards another common endpoint with regards to the 30 different mutagenic profiles.”

7) The fact that the tumors did not identify any changes is interesting and needs to be further addressed.

As stated in subsection “In vivo studies confirm the fully transformed state of cancer cell exosome-initiated cells”:

“Histological analysis confirmed that the tumors are fibrosarcomas, as was expected because cells used in the CTA are NIH/3T3 cells from a mesenchymal origin (Figure 3—figure supplement 1D).”

Reviewer #3:

This is a potentially interesting study that clearly shows that (1) Cancer-derived exosomes effectively initiate malignant transformation in mouse NIH3T3 cells in vitro; (2) The transformed NIH3T3 cells (capable of foci formation) form subQ fibrosarcomas; and (3) not unexpectedly, exosome-induced tumor initiation is related to defective DNA repair, however, the repair defects are random (not caused by one specific exosome-associated oncogene).

The manuscript is well executed, and provides multiple controls, such as additional promoters, transposition of initiator and promoter in two-step carcinogenesis assay, etcetera. However, the study its limited to investigation of NH3T3 transformation and these cells are notoriously easy to transform. Also, the biological relevance of the study is somewhat questionable and critical issues remain unresolve

First, could a similar process occur in vivo, in mice or humans? If so, an individual with a primary pancreatic tumor should have a higher propensity to form other primary tumor(s). Is that the case?

These are interesting questions raised by the reviewer and we discussed this as a fascinating idea based on a recent study by Felsentein and colleagues where they describe a finding in which a portion of co-occurring IPMNs in PDAC patients appeared to be genetically unrelated, meaning they shared no mutations in the assayed genes (Discussion section):

“A recent study by Felsentein et al. described an interesting finding in which a portion of co-occurring IPMNs in PDAC patients appeared to be genetically unrelated, meaning they shared no mutations in the assayed genes (42). This elicits a fascinating question as to whether cells in a primary tumor could be derived from independent transformation events as opposed to exclusive clonal events (42). Potentially, exosomes secreted by the primary tumor could orchestrate such events. Additional studies towards understanding the cancer cell exosome mediated initiation are needed to address such questions, specifically, investigating the initiation capacity of cancer cell exosomes in more relevant cells like human epithelial cells.”

In addition, to further address questions asked by this reviewer, we have incorporated discussion on the role that exosomes have been previously shown to play in cell transformation in the Discussion section:

“Recent studies demonstrate that cancer cell exosomes can in fact participate in cell transformation, however the details on how these exosomes contribute to transformation remained elusive. For example, Dai et al. showed that exosomes participated in the transformation process of normal cells after cells were treated with arsenite (15). Another study described a transfer of malignant traits to BRCA1 deficient human fibroblast when treated with either cancer patient sera or isolated cancer cell exosomes, leading to malignant transformation (16, 17). In addition, exosomes in breast cancer patient sera have been shown to promote nontumorigenic epithelial cells to form tumors in a Dicer-dependent manner (11). In each of the aforementioned studies, non-transformed cells were either pre-exposed to transforming agents or treated with cancer patient sera or medium from cultured cancer cells. Interestingly, Antonyak et al. has demonstrated that sustained treatment of NIH/3T3 cells with breast cancer or glioma cell-derived microvesicles has the ability to induce transformation of recipient cells (18). Contrastingly, in our studies, when cells were treated only with pancreatic cancer cell exosomes, increased cell transformation was not observed. While the specific details and effects on cell transformation attributed to exosomes vary between the studies, both demonstrate that exosomes are indeed playing a role in cell transformation.”

Do such individuals have a chance of exposure to promoters?

Most likely everyone is exposed to such factors, as many known promoters are made by cells. Normal cells are able to use intact signaling machinery and defense mechanisms to respond to such factors. These would include growth factors, oxidative-stress, inflammation, etc. Both oxidative stress and inflammation are associated with pancreatitis and pancreatic cancer and could function as an intrinsic promoter.

Second, could this phenomenon be extended to other tumor types (not PDAC)? And last, would exosome-initiated tumors be exclusively fibrosarcomas? Other tumor types? Random tumors? Would PDAC exosomes act as tumor initiating agent in pancreatic epithelium?

Explanation and reasoning behind our use of the classic 2-stage cell transformation assay was added to the Introduction section:

“Malignant transformation of a normal cell occurs in a stepwise fashion. Point mutations in the genome can result in the reprogramming of a normal cell to a less differentiated state that is receptive to additional genetic alterations resulting in uncontrolled growth and ultimately cancer. The classic 2-stage in vitro cell transformation assay (CTA) is a tiered system for transformation that was created for screening potential carcinogenic factors (21-23). In this system, cells are first treated with a suspected carcinogen, called an initiator, such as the genotoxic carcinogen 3-MCA (3-methylcholanthrene). 3-MCA introduces random genetic changes in a pool of normal cells. Subsequently, these initiated cells are exposed to a promoter, such as TPA (12-O-tetradecanoylphorbol 13-acetate), which enhances proliferation in the initiated cells selectively, thus driving malignant transformation of the cells. The resulting transformed cells are observed as foci on a cell culture plate (23, 24). This reductionist approach provides sensitivity in detecting a wider range of initiating agents that may not show obvious transforming activity without a promoter (23). Using this assay as a model system for malignant cell transformation, we assessed whether or not cancer cell-derived exosomes could affect and/or potentially drive the transformation of a normal cell.

The results presented herein provide a detailed analysis of a previously unidentified molecular function of cancer cell exosomes for malignant cell transformation.”

It will be very interesting to test this function in other cell types, and will be done in future studies. As the optimization for another assay (2-3 months/assay condition) to be equally reliable as the CTA with 3T3-cells will require over a year, we think this experiment is beyond the scope of this paper.

These questions could be answered only by using PDAC exosomes in a classical two-step carcinogenesis assay, as opposed to in vitro transformation assay followed by tumorigenesis assay. Similar experiments demonstrating transformation events due to the transfer of exosomes' contents to the bystander cells have already been performed using exposures isolated from the tumor cell lines or from the sera of cancer patients (please see reference below). Most of these experiment are performed as a one step process, which also contradicts the conclusions from the present analysis.

Actually, other studies are done in the context of another factor. We discussed this in the Discussion section:

“Recent studies demonstrate that cancer cell exosomes can in fact participate in cell transformation, however the details on how these exosomes contribute to transformation remained elusive. For example, Dai et al. showed that exosomes participated in the transformation process of normal cells after cells were treated with arsenite (15). Another study described a transfer of malignant traits to BRCA1 deficient human fibroblast when treated with either cancer patient sera or isolated cancer cell exosomes, leading to malignant transformation (16, 17). In addition, exosomes in breast cancer patient sera have been shown to promote nontumorigenic epithelial cells to form tumors in a Dicer-dependent manner (11). In each of the aforementioned studies, non-transformed cells were either pre-exposed to transforming agents or treated with cancer patient sera or medium from cultured cancer cells. Interestingly, Antonyak et al. has demonstrated that sustained treatment of NIH/3T3 cells with breast cancer or glioma cell-derived microvesicles has the ability to induce transformation of recipient cells (18). Contrastingly, in our studies, when cells were treated only with pancreatic cancer cell exosomes, increased cell transformation was not observed. While the specific details and effects on cell transformation attributed to exosomes vary between the studies, both demonstrate that exosomes are indeed playing a role in cell transformation.”

Finally, the potential ability of exosomes to promote tumor growth by virtue of transforming host cells could be tested using fluorescent mice (GFP, RFP etc) as donors or acceptors of exosome-producing NIH-3T3 cells. The appearance of fluorescence tagged cells in a non-tagged exotic tumor formed by exosome-transformed NIH-3T3 cells, whose origin could be ascertained using other fluorescent makers would clearly ascertain such recruitment.

We appreciate these suggestions made by the reviewer and will take these into consideration in future studies but we think that they are beyond the scope of this paper.

Regrettably, this potentially interesting study is insufficiently novel to be published in eLife in its present form, given limited scope of analyses performed and potential low relevance of the results.

We specifically used the 3T3 assay to ask a precise mechanistic question because this would allow us to use a reductionist approach to define a very elusive question. Do cancer cell or normal cell exosomes act as an initiator, promoter or both in cell transformation? We utilized the 2-stage CTA as a reliable model assay to answer this question. In addition, the 2-stage CTA could be performed with NIH/3T3 cells using an established chemical initiator (MCA) and promoters (TPA and CdCl2).

In addition, as mentioned in the title, summation and manuscript, we have identified, for the first time, that cancer cell exosomes can act as classic initiators in the cell transformation. Our studies provide mechanistic insight by demonstrating that cancer cell exosomes, like other classic chemical initiators, induce random mutations into cells. Cells from this pool of randomly mutated cells can then be driven to transformation by a classic promoter.

In favor of the potential relevance of this study, 18% of co-occurring IPMN and ductal adenocarcinomas were likely independent, suggesting that the carcinoma arose from an independent precursor. By contrast, all colloid carcinomas were likely related to their associated IPMNs. However, to link this process to exosome-dependent tumor initiation, one has to analyze exosomes from patients from IPMNs and their potential capacity to induce pancreatic tumors. (Felsenstein et al). However, the present study is extremely preliminary in nature and does not unequivocally address such a possibility. Further experiments are needed to demonstrate the tumor-initiating capacity of exosomes from the neoplastic foci.

https://www.nature.com/articles/s41419-018-0485-1.pdf (Daie et all, Cell Death and Disease (2018)9:454); https://jeccr.biomedcentral.com/track/pdf/10.1186/s13046-017-0587-0 (Abdouh et al., Journal of Experimental & Clinical Cancer Research (2017) 36:113);

http://www.pnas.org/content/108/12/4852 (Antonyak et al., 2011. PNAS 108 (12) 4852-4857); https://doi.org/10.1016/j.bbrc.2014.07.109 (Lee et al., BBRC Volume 451, Issue 2, 22 August 2014, Pages 295-301

http://cancerres.aacrjournals.org/content/77/21/5808.full-text.pdf (Figueroa et al. Cancer Res. 2017 Nov 1;77(21):5808-5819. doi

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4254633/pdf/nihms631122.pdf (Melo et al., Cancer Cell. 2014 November 10; 26(5): 707-721)

Additional data files and statistical comments:

The experiments as presented are performed rigorously, with appropriate statistical analysis…

Thank you for acknowledging the rigor in our work. We have added more detailed information to our Materials and methods to explain our protocols used for bioinformatic analysis.

…however the conclusions that could be drawn are limited and insufficiently novel.

As mentioned in the title, and described in the summation and manuscript, we have identified, for the first time, that cancer cell exosomes can act as classic initiators in the cell transformation. Our studies provide mechanistic insight by demonstrating that cancer cell exosomes, like other classic chemical initiators, induce random mutations into cells. Cells from this pool of randomly mutated cells can then be driven to transformation by a classic promoter.

[Editors’ note: the evaluation of the paper after the first round of revision follows.]

While you have performed a few additional experiments that strengthen the manuscript, the reviewers agree that many of the original criticisms remain unaddressed. As you are aware, in this eLife trial we are committed to ultimately publishing the paper should you choose to proceed with it. However, all of the reviewer and editorial concerns about the validity of these findings will be published in parallel with the paper.

There remain several major areas of concern:

1) The methods used to isolate the exosomes.

Randy Schekman specifically commented upon the new data as follows:

"I have examined the additional data relevant to the concern I raised about the crude fraction they use to claim an exosome activity. As far as I can tell this consists of one data point with a peak fraction of buoyant membranes assayed at one concentration. This would not pass muster by any standard biochemical analysis. I recommend assays across the peak fraction of exosomal marker proteins and a quantitative titration of activity vs. protein concentration in the peak fraction in comparison to the crude starting material. The expectation is that if this activity tracks with exosomes, the specific activity in their biological assay should have been enriched by the buoyant density separation. Without this it is premature to conclude that the activity is associated with exosomes.”

The experiments previously requested by this editor in initial reviews were to perform “further fractionation of the relevant sedimentable species” in order to ensure that exosomes were properly isolated and purified from contaminating material. To address this concern, exosomes were further purified on a sucrose density gradient and resulting fractions were blotted for common exosomal markers CD63 and Alix, using the protocol recommended by this editor. Exosome marker proteins were only found in fractions 3 and 4, consistent with previous results. Additionally, nanoparticle tracking analysis of ‘fraction 3’ isolated exosomes clearly show enrichment in particles sized <100nm compared to analysis of ‘crude’ exosomes. After protein concentration of the fractions was determined, Fraction 3 was tested in the cell transformation assay at a concentration of 80 ng/mL (or 7x107 particles/mL) to stay consistent with crude exosome assays. This concentration of exosomes was determined based on dose response studies already performed and described in the paper. When Fraction 3 exosomes were used as an “initiator,” results in Figure 2—figure supplement 1B show that the average foci/well formed is comparable to results from crude exosomes. These experiments demonstrate that both the initial population of exosomes and the further purified fraction of exosomes requested by the editor function as an initiator in the assay.

2) The notion that exosomes are initiators.

Several of the reviewers raised concerns about using 3T3 cells as the sole evidence of initiation activity. This was not addressed experimentally, but instead the authors state that "It will be very interesting to test this function in other cell types, and will be done in future studies.” Despite this being a major concern, this is still left unaddressed.

The assay used in this manuscript is a classic assay that is well accepted for discovery of both initiators and promoters of cellular transformation. This in vitro cell transformation assay (CTA) is proposed as an alternative to in vivo carcinogenicity testing. This assay is reported in the list of accepted methods for REACH (Reg. EC 440/2008). (1) Balb/c and NIH/3T3 cells are commonly used cell lines in the Cell Transformation Assay (CTA) and Focus Formation Assay (FFA) (2) which are assays that provide a straightforward method to assess the transforming potential of an oncogene or biological component. These assays represent the most well-known CTAs and are regarded as a reliable tool to screen biological components, single chemicals or mixtures for carcinogenicity prediction. In our experiments we used the two-stage cell transformation system which according to Sakai ‘is utilized for measuring tumor promoting activities and also for enhancing the sensitivity of cells to the weak initiators the standard assay is unable to detect’ (3, 4). Accordingly, by using this assay the end result produces a reproducible phenotype of foci formation. Rigorous standards for foci scoring have been previously established and were used for our studies (5). In this paper, we have used this assay with relevant positive and negative controls and established in a statistically significant and scientific manner that cancer cell exosomes can function as an initiator within the context of the assay.

Unfortunately, the same assay does not result in the same phenotype when using human normal epithelial cell lines. Human cells are known to be more resistant to cell transformation than rodent cells (6, 7). This known limitation should not preclude use of the rodent CTA. Currently, there is no other reliable system for the analysis of an initiation and promotion of cell transformation and it will take significant time to develop a new and reliable 8-week assay. We are in the process of designing a new assay with human cell lines to study the initiator and promoter activity of known compounds and exosomes, but this new assay is beyond the scope of this study.

As the 3T3 assay is well accepted and rigorous “to screen biological components, single chemicals or mixtures for carcinogenicity prediction”, we think this assay is sufficient for this discovery.

Why this is important is related to the new data in Figure 3B, Figure 3—figure supplement 2A, they show that the background foci actually do form tumors, albeit at a slower rate and with lower efficiency than the exosome treated 3T3 foci. But the fact that the background foci do indeed form tumors actually argues against a pure initiator activity – it seems to accelerate what is already happening in the background foci anyway. Thus, a central argument that these exosomes are true initiators does not seem fully accurate. In the absence of other cell types other than 3T3s, it is hard to state this definitively outside of this one context.

The reviewers initially asked us to test whether background control foci would form tumors when injected into mice. We did these experiments and results showed that background foci from untreated cells could form tumors when injected into mice, albeit these tumors were observed to grow at a less aggressive rate. These background foci are present in all experiments, due to the fact that NIH/3T3 cells are susceptible to transformation. It has been shown that when NIH/3T3 cells are maintained as confluent, non-growing cultures without passage, or injected into mice in high concentrations, they may undergo spontaneous cell transformation (8). It is therefore not unexpected that we observed tumor growth after injection of transformed background foci into mice. The probability to spontaneously transform is also increased if cells have been passaged several times (n>4) (9). Maintaining the same passaged cells throughout the experiment is an important factor for obtaining reliable results (9). Each of our experiments were performed with the sub-confluent cells passaged less than four times.

The presence of these background foci also does not diminish the role of initiators and promoters in cell transformation. Using strict focus scoring based on previously established criteria, we established a constant level of background foci through all experiments; in each experiment a small number of cells transformed independent of any treatment, as has been observed in previous studies (10). This does not argue against the initiator activity of exosomes as the reviewer suggests, because only in experiments where cells were treated with a functional initiator and a functional promoter, did we observe a significant increase in transformation above background. This same increase above background is observed for the established chemical initiator, MCA. Finally, in keeping with the definition of an initiator, we demonstrate that initiators (cancer cell exosomes or MCA) can work with multiple promotors (TPA or CdCl2).

References:

1. M. G. Mascolo et al., BALB/c 3T3 cell transformation assay for the prediction of carcinogenic potential of chemicals and environmental mixtures. Toxicology in vitro: an international journal published in association with BIBRA24, 1292-1300 (2010).

2. in Encyclopedia of Cancer, M. Schwab, Ed. (Springer Berlin Heidelberg, Berlin, Heidelberg, 2009), pp. 2081-2081.

3. A. Sakai, BALB/c 3t3 cell transformation assays for the assessment of chemical carcinogenicity. Alternatives to Animal Testing and Experimentation14, 367-373 (2007).

4. A. Sakai, M. Sato, Improvement of carcinogen identification in BALB/3T3 cell transformation by application of a 2-stage method. Mutation research214, 285-296 (1989).

5. K. Sasaki et al., Photo catalogue for the classification of foci in the BALB/c 3T3 cell transformation assay. Mutation research744, 42-53 (2012).

6. T. Akagi, K. Sasai, H. Hanafusa, Refractory nature of normal human diploid fibroblasts with respect to oncogene-mediated transformation. Proceedings of the National Academy of Sciences of the United States of America100, 13567-13572 (2003).

7. S. Creton et al., Cell transformation assays for prediction of carcinogenic potential: state of the science and future research needs. Mutagenesis27, 93-101 (2012).

8. K. Xu, H. Rubin, Cell transformation as aberrant differentiation: Environmentslly dependent spontaneous transformation of NIH 3T3 cells. Cell Research1, 197 (1990).

9. K. Sasaki, A. Huk, N. E. Yamani, N. Tanaka, M. Dusinska, in Genotoxicity and DNA Repair: A Practical Approach, L. M. Sierra, I. Gaivão, Eds. (Springer New York, New York, NY, 2014), pp. 343-362.

10. H. Maeshima, K. Ohno, Y. Tanaka-Azuma, S. Nakano, T. Yamada, Identification of tumor promotion marker genes for predicting tumor promoting potential of chemicals in BALB/c 3T3 cells. Toxicology in vitro: an international journal published in association with BIBRA23, 148-157 (2009).

3) The proteomic analysis.

While the authors make it clear how they performed the analysis, it is still confusing to state "However, proteins found in transformed cells shown in Figure 4 exhibit more divergence from control cells" Divergence based on what statistical measure? Venn diagrams are not statistical methods, and it is not clear that the PANTHER data shows true statistical enrichment in the exosome treated cells.

We have clarified the text pertaining to how the proteomic data is interpreted. Venn diagrams are used to directly compare proteins found in each sample and highlight differences between proteins identified in transformed foci with untreated cells.

4) Exome analysis.

In the rebuttal, the authors state that "we feel confident that the method we used is sufficient to obtain reliable results". Yet the request to analyze this data with additional pipelines and more stringent statistics was not performed. There is no orthogonal validation of this pipeline to ensure that the results are accurate.

We used the Strelka2 somatic variation pipeline as previously published and stated in Materials and methods (Kim, S., Scheffler, K. et al. (2018) Strelka2: fast and accurate calling of germline and somatic variants. Nature Methods, 15, 591-594. doi:10.1038/s41592-018-0051-x). We have clarified the text in the Materials and methods to accurately reflect this in the last three paragraphs of the Materials and methods section.

In looking at their more detailed methods, it is now clear that they are comparing their exomes to the human reference genomes (GRCm38), so that the variants being called are actually just variants that differ from this reference genome, and not the parental 3T3 cells. How do you know that most of the called variants are not just called SNPs or baseline point mutations between 3T3 cells and the human reference genome? Why not compare the exome data to the parental 3T3 cells rather than the human reference genome?

The sequencing data is being compared to mutations found in the parental NIH/3T3 cells. As NIH/3T3 are mouse cells, we compare our data with both the mouse reference genome and then with the original NIH/3T3 cells. As described in Materials and methods, we first align the raw sequencing data to the mouse reference genome (GRCm38). We have clarified the text in the Materials and methods to accurately reflect this in the last three paragraphs of the Materials and methods section.

The authors now provide additional data, using the same pipelines, on bulk exome data on both untreated and TPA treated foci in Figures 5 and 6. In Figure 5—figure supplement 1, they essentially show that the number of mutations in the untreated or TPA treated cells is similar to that in the MCA/TPA or ExC/TPA foci. But if the basic idea is that exosomes act as initiators through increases in mutation rate, then why are the number of mutations essentially the same.

The last statement is incorrect as it compares two unrelated events. The first part of this statement, “But if the basic idea is that exosomes act as initiators through increases in mutation rate”, is correct and we propose that cancer cell exosomes like other initiators (such as MCA) act by causing mutations in the initiator step of the transformation process. This provides an increase in the number of mutated cells that can be used by a promoter to produce more foci than are observed in control wells.

The second part of the statement “then why are the number of mutations essentially the same” is unrelated to the first. The number of mutations in the cells from foci is due to the transformation process whereby a normal cell becomes a transformed cell that has accumulated many mutations of over time. For all the foci, COSMIC mutational profiles (20,15 or 3) involving some type of DNA damage were revealed by MutaGene analysis. As a result, the number of mutations is expected to be large. Additionally, the number of mutations depicted in the Venn diagrams reflects the full set of variants found in each sample, prior to filtering for loss of function or missense mutations. Further filtering of the mutations for loss of function and missense mutations was used when analyzing the specific mutations found in the MMR associated genes (Figure 6—figure supplement 1 and 2) and the 190 oncogenes (Figure 5—figure supplement 1-source data 1).

Furthermore, in the data in Figure 6, you show that the signature 20 is higher in the MCA/TPA or ExC/TPA samples compared to untreated/TPA, but this just looks like a modest increase, and there is no statistical analysis to say whether these are really different.

The data in Figure 6 represents the relative contribution of the pan-cancer derived COSMIC signatures to the mutational profiles of each foci. Contribution of signatures was calculated using MutaGene, the query profile is decomposed into the precomputed sets for COSMIC to determine the exposure of a given sample to different mutation processes (Figure 6—source data 1). Discussion of these profiles is found in the Discussion.

How do you confirm that you are really enriching for this signature in this dataset other than a heatmap which looks to be so.

Respectfully, this is a correct assessment as we cannot definitively confirm or prove the hypothesis that TPA when acting on initiated cells will cause mutations in mismatch repair genes. We therefore have changed the text to read:

“These observations support the proposal that when TPA is used as a promoter (to drive proliferation) in combination with a functional initiator (an entity that creates a population of randomly mutated cells), the molecular path of cells towards tumorigenesis may not be random. Rather, TPA may drive a path in a subset of randomly mutated cells that accommodates a course towards COSMIC signatures 20 or 15. Furthermore, future studies might reveal that another promoter acting on initiator treated cells might lead cells towards another common endpoint with regards to the 30 different signature profiles.”

Finally, it is not clear that bulk sequencing of this population of highly mutated cells is appropriate. If we consider that each cell could have many different mutations, the more appropriate experiment is to create single cell clones out of the transformed foci, let that grow up, and then perform the exome analysis on these isolated clones. While this is an expensive and ultimately very time consuming method, it would ultimately be much more informative in understanding how the mutations occur in each cell, rather than a population of cells which may or may not have defects is mismatch repair.

One cannot expect to have a clonal population when a genome encodes a mutation in a mismatch repair gene. This was previously demonstrated:

Orth, K., Hung, J., Gazdar, A., Mathis, M., Bowcock, A., & Sambrook, J. Ovarian tumors display persistent microsatellite instability caused by mutation in the mismatch repair gene hMSH-2. Cold Spring Harbor Symp. Quant. Biology (1994) 59, 349-56.

5) The role of mismatch repair.

While we recognize that adding in single mismatch repair genes may not rescue the phenotypes, at the same time the authors are making a very bold claim that exosomes specifically lead to MMR defects. This is perhaps the most intriguing idea in the paper, but without any experiments to mechanistically dissect this it remains correlative.

We do not claim that exosomes specifically lead to mismatch repair defects.

We stated that an initiator (that causes random mutations) in combination with the promoter TPA was observed to result in a mismatch repair phenotype as indicated by our MutaGene analysis.

https://doi.org/10.7554/eLife.40226.032

Article and author information

Author details

  1. Karoliina Stefanius

    1. Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, United States
    2. Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9456-1954
  2. Kelly Servage

    1. Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, United States
    2. Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7183-2865
  3. Marcela de Souza Santos

    Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Data curation, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Hillery Fields Gray

    1. Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, United States
    2. Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Jason E Toombs

    Division of Surgical Oncology, Department of Surgery, Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  6. Suneeta Chimalapati

    1. Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, United States
    2. Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  7. Min S Kim

    Department of Bioinformatics, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Data curation, Software, Validation, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  8. Venkat S Malladi

    Department of Bioinformatics, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Data curation, Formal analysis, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  9. Rolf Brekken

    1. Division of Surgical Oncology, Department of Surgery, Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center, Dallas, United States
    2. Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2704-2377
  10. Kim Orth

    1. Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, United States
    2. Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, United States
    3. Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Investigation, Writing—original draft, Writing—review and editing
    For correspondence
    kim.orth@utsouthwestern.edu
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0678-7620

Funding

National Institutes of Health (GM115188)

  • Kim Orth

Once Upon A Time Foundation

  • Kim Orth

Welch Foundation (I-1561)

  • Kim Orth

National Institutes of Health (CA192381)

  • Rolf Brekken

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

Acknowledgements

We thank members of the Orth lab for their helpful discussions and advice. We thank John Minna for his critical assessments and support. We thank Diego Castrillon for his expert advice on tissue pathology and the UTSW Electron microscopy Core Facility. In memory and thanks to our friend and colleague Dr. Alfred G Gilman. 

Ethics

Animal experimentation: This study was performed in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and according an approved UT Southwestern institutional animal care and use committee (IACUC) protocol (APN 2016-101732).

Senior Editor

  1. Jeffrey Settleman, Calico Life Sciences, United States

Reviewing Editor

  1. Richard M White, Memorial Sloan Kettering Cancer Center, United States

Publication history

  1. Received: July 19, 2018
  2. Accepted: May 8, 2019
  3. Version of Record published: May 28, 2019 (version 1)

Copyright

© 2019, Stefanius 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

  • 2,405
    Page views
  • 262
    Downloads
  • 2
    Citations

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

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)

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

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

Further reading

    1. Cancer Biology
    2. Human Biology and Medicine
    Chen Farhy et al.
    Research Article Updated
    1. Cancer Biology
    2. Computational and Systems Biology
    Adam C Palmer et al.
    Research Article Updated