1. Cancer Biology
  2. Human Biology and Medicine
Download icon

FGF2-FGFR1 signaling regulates release of Leukemia-Protective exosomes from bone marrow stromal cells

  • Cited 0
  • Views 1,033
  • Annotations
Cite this article as: eLife 2019;8:e40033 doi: 10.7554/eLife.40033

Abstract

Protective signaling from the leukemia microenvironment leads to leukemia cell persistence, development of resistance, and disease relapse. Here, we demonstrate that fibroblast growth factor 2 (FGF2) from bone marrow stromal cells is secreted in exosomes, which are subsequently endocytosed by leukemia cells, and protect leukemia cells from tyrosine kinase inhibitors (TKIs). Expression of FGF2 and its receptor, FGFR1, are both increased in a subset of stromal cell lines and primary AML stroma; and increased FGF2/FGFR1 signaling is associated with increased exosome secretion. FGFR inhibition (or gene silencing) interrupts stromal autocrine growth and significantly decreases secretion of FGF2-containing exosomes, resulting in less stromal protection of leukemia cells. Likewise, Fgf2 -/- mice transplanted with retroviral BCR-ABL leukemia survive significantly longer than their +/+ counterparts when treated with TKI. Thus, inhibition of FGFR can modulate stromal function, reduce exosome secretion, and may be a therapeutic option to overcome resistance to TKIs.

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 all the issues have been addressed (see decision letter).

https://doi.org/10.7554/eLife.40033.001

eLife digest

Leukemias are cancers of white blood cells. The cells grow and divide rapidly, often because of mutations in proteins called kinases. Since the kinase mutations do not occur in healthy cells, they provide a good target for anti-leukemia drugs. Several such kinase inhibitors are effective at treating leukemia patients. However, most leukemia cells develop ways to resist the effects of the kinase inhibitors over time, leading to relapses of the disease.

One way that leukemia cells resist kinase inhibitors is by taking advantage of signals coming from supportive cells, known as stromal cells, in the bone marrow. When patients are treated with kinase inhibitors, the bone marrow stromal cells produce more of a signaling protein called FGF2. The leukemia cells then use FGF2 to survive the effects of the kinase inhibitors.

It was not clear how the FGF2 signal reaches the leukemia cells from the bone marrow stromal cells. Now, using biochemical techniques, Javidi-Sharifi, Martinez et al. show that bone marrow stromal cells package FGF2 into small compartments called exosomes. The stromal cells release the exosomes into the bone marrow, and the leukemia cells then engulf and internalize the exosomes. Leukemia cells that had taken up FGF2 in this way were better able to survive kinase inhibitor treatment than leukemia cells that had not.

Javidi-Sharifi, Martinez et al. also observed that FGF2 also affects the bone marrow stromal cells themselves, causing them to grow faster, produce more FGF2 and release more exosomes. Blocking the effects of FGF2 on the stromal cells slowed their growth and caused fewer exosomes to be released. In addition, mice whose bone marrow stromal cells could not produce FGF2 survived leukemia for longer than mice whose stromal cells provided protective FGF2 in exosomes to leukemia cells. This suggests that taking advantage of drugs that prevent bone marrow stromal cells from releasing FGF2 in exosomes might improve treatments for leukemia. Further research will be needed to confirm whether this strategy would be effective in humans.

https://doi.org/10.7554/eLife.40033.002

Introduction

TKIs have revolutionized the treatment of CML and have shown promise in AML, however development of resistance remains a problem. In CML, resistance develops in a minority of patients, and is most often caused by resistance mutations. However, some patients still develop resistance in the absence of known resistance mutations. In contrast, development of resistance in AML is the norm. Inhibitors of mutated FLT3, which is present in about 30% of AML patients, are initially quite efficacious (Smith et al., 2012). However, resistance to FLT3 kinase inhibitors in AML typically develops within a few months. In some cases, resistance is cell-intrinsic and due to secondary mutations in the activating loop of FLT3 that prevent drug binding (Weisberg et al., 2009), however, resistance still develops in the absence of these mutations. Within the marrow microenvironment, leukemia cell survival can be mediated by extrinsic ligands that activate alternative survival pathways (Smith et al., 2017; Ghiaur and Levis, 2017; Wilson et al., 2012) and over time can lead to development of intrinsic resistance mutations (Wilson et al., 2012; Traer et al., 2012).

Bone marrow stromal cells provide a supportive structure and secrete cytokines that contribute to the normal hematopoietic stem cell niche, but can also protect leukemic cells from therapy (Colmone et al., 2008; Ayala et al., 2009). Initial studies into the mechanisms of resistance utilized normal marrow stroma (Manshouri et al., 2011), but the stroma can be altered by leukemia, in a manner similar to development of cancer associated fibroblasts in solid tumors (Paggetti et al., 2015; Huang et al., 2015; Huan et al., 2015). We found that fibroblast growth factor 2 (FGF2) expression is increased in marrow stromal cells during tyrosine kinase inhibitor (TKI) therapy and protects leukemia cells (Ware et al., 2013; Traer et al., 2014; Javidi-Sharifi et al., 2015; Traer et al., 2016). FGF2 has also been shown to be essential for stress hematopoiesis after chemotherapy (Itkin et al., 2012; Zhao et al., 2012), suggesting that leukemia cells can hijack a normal marrow stress response for their own survival.

Despite its important roles in physiology and pathology, several aspects of FGF2 biology remain poorly understood. FGF2 does not have a signal peptide and is not secreted through the canonical secretory pathway. Alternative mechanisms for secretion have been proposed, but how FGF2 is conveyed between two cells remains unclear (Steringer et al., 2015; Zacherl et al., 2015). Additionally, while recombinant FGF2 directly stimulates myeloid colony formation (Berardi et al., 1995), there are also reports suggesting that FGF2 can indirectly regulate hematopoiesis by stimulating stromal cells to produce cytokines (Avraham et al., 1994).

We discovered that FGF2 is largely secreted in extracellular vesicles (ECVs) and exosomes from bone marrow stromal cells. ECVs are able to protect leukemia cells from the effects of TKI therapy. Furthermore, autocrine FGF2-FGFR1 activation in marrow stromal cells increases the secretion of FGF2-laden exosomes, indicating that exosome secretion is regulated in part by FGF2-FGFR1 signaling. Inhibition of FGFR1 can reverse this protective stroma-leukemia interaction and restore leukemia cell TKI sensitivity in the marrow niche.

Results

Stromal cell ECVs protect leukemia cells from TKI therapy

The human stromal cell line HS-5 expresses abundant FGF2, in addition to other soluble cytokines such as IL-5, IL-8 and HGF (Roecklein and Torok-Storb, 1995), and conditioned media (CM) from HS-5 is highly protective of leukemia cell lines. HS-5 CM was ultracentrifuged at 100,000 g to separate soluble proteins (supernatant, S100) from ECVs and larger macromolecules (pellet, P100). We compared the protective effect of unfractionated CM, S100, and P100 fractions on the viability of two leukemia cell lines: MOLM14 (FLT3 ITD+ AML) and K562 (CML), in the presence of their respective TKIs, quizartinib (AC220, a highly selective and potent inhibitor [Zarrinkar et al., 2009]) and imatinib (Figure 1A and B). The protective capacity of the S100 fraction was less than unfractionated CM, and protection was enriched in the concentrated P100 ECV fraction (Figure 1), indicating that a substantial protective component of HS-5 CM is mediated by ECVs. A more extensive profiling of protection is also shown in Figure 1—figure supplement 1.

Figure 1 with 3 supplements see all
Extracellular vesicles (ECVs) secreted by HS-5 cells are internalized by MOLM14 and K562 cells and protect from treatment with AC220 or imatinib, respectively.

HS-5 conditioned media (CM) was collected and separated by ultracentrifugation at 100,000 g into a supernatant (S100) and pellet (P100) fraction containing ECVs. These fractions were incubated with (A) K562 cells ± 1 μM imatinib, or (B) MOLM14 cells ± 10 nM AC220, and viability measured by MTS assay after 48 hr. Values were normalized to respective untreated condition. All wells were plated in triplicate and error bars indicate standard deviation. RPMI is the media control. p values are indicated by *<0.05, **<0.005, and ***=0.0007. (C), MOLM14 and K562 cells were stained with DiO (green) tracer, washed, and immobilized on Poly-D-Lysine coated chamber slides. HS-5 P100 fraction was stained with DiI (red) tracer and added to the cells for a 24 hr incubation. Slides were stained with DAPI (blue) and imaged by confocal fluorescent microscopy. A movie of the z-stack images is included in Supplemental data.

https://doi.org/10.7554/eLife.40033.003

To determine if ECVs produced by HS-5 cells are internalized by K562 and MOLM14 leukemia cells, K562 and MOLM14 cells were stained with a green lipophilic tracer (DiO) and incubated with HS-5 ECVs stained with a red lipophilic tracer (DiI). Analysis by confocal microscopy showed that ECVs are indeed internalized by leukemia cells, although the exact mechanism of internalization is still under investigation (Figure 1C and Figure 1—video 1 and 2).

FGF2 is contained in stromal cell ECVs and exosomes

FGF2 is highly expressed in the HS-5 stromal cell line but the related HS-27 expresses little FGF2 (Figure 2ATraer et al., 2016). We analyzed FGF2 in S100 and P100 fractions of both HS-5 and HS-27 by immunoblot (Figure 2B). Little FGF2 was detected in the soluble protein fraction (S100), but FGF2 was enriched in ECVs (P100). Washing the ultracentrifuge tube with detergent liberated even more FGF2 (detergent wash P100), due to ECVs adhering to the ultracentrifuge tube. To compare FGF2 to other soluble cytokines, HS-5 CM was ultracentrifuged into S100 and ECVs, cytokines quantified by Luminex multiplex assay, and normalized to unfractionated CM (Figure 2C). Pelleted ECVs were resuspended in 10% of the original CM volume, and the P100 bars in Figure 2B thus represent a 10-fold enrichment, although as shown in Figure 2B not all ECVs can be liberated from the ultracentrifuge tube. FGF2 was uniquely enriched in ECVs, whereas soluble cytokines such as stem cell factor, interleukin (IL)−6, IL-8, etc. were found primarily in the S100 fraction.

FGF2 is enriched in exosomes from HS-5 bone marrow stromal cells.

(A) Immunoblot of FGFR1, FGF2 and actin in HS-5 and HS-27 whole cell lysates. (B) HS-5 and HS-27 CM were ultracentrifuged at 100,000 g for 2 hr at four degrees C. CM, soluble protein (S100), and ECV (P100) fractions were collected and analyzed by immunoblot, using 10, 50, and 100 ng/ml recombinant FGF2 for comparison. The ultracentrifuge tube was also washed with detergent to remove adherent ECVs and material (detergent wash P100). (C) HS-5 CM, S100 and P100 fractions (concentrated ~10 fold compared to HS-5 CM) were solubilized in 0.1% NP-40 and analyzed by cytokine multiplex ELISA (Luminex). The S100 and P100 fractions were normalized to CM. (D) The HS-5 P100 fraction (starting material, or SM) was further fractionated on a sucrose step-gradient. Sucrose layer interfaces (0–7.5%, 7.5–15%, 15–30%, 30–45%, and 45%-pellet) were collected, lysed and analyzed by immunoblot with antibodies against the exosomal marker CD9, FGF2, and cytoplasmic marker actin. (E) HS-5 and HS-27 ECVs (P100), recombinant FGF2, and HS-5 cells were exposed to proteinase K and analyzed by immunoblot. (F) HS-5 exosomes were isolated by sucrose step-gradient (see panel D) and then exposed to proteinase K with or without detergent (0.1% Triton X-100, used to dissolve the lipid membrane). Samples were subjected to immunoblot analysis using antibodies against tsg101, CD9 and FGF2. The * indicates degraded FGF2 after partial proteinase K digestion.

https://doi.org/10.7554/eLife.40033.007

HS-5 ECVs were further separated into microvesicles, exosomes, and insoluble extracellular matrix proteins (ECM) using a sucrose step-gradient to separate by density. FGF2 and cell compartment-specific molecular markers were probed by immunoblot (Figure 2D). FGF2 was most highly enriched in the 15–30% sucrose interface, which also contained the exosome-specific marker CD9.

To determine if FGF2 was bound to the outside of ECVS, or contained within ECVs, proteinase K was used to digest proteins not enclosed by lipid membrane. Recombinant FGF2, HS-5 or HS-27 ECVs, and intact HS-5 cells were incubated with proteinase K and probed for FGF2 by immunoblot (Figure 2E). Recombinant FGF2 was completely degraded by proteinase K (* indicates degraded fragments) but intact FGF2 was detected in both HS-5 ECVs and control HS-5 cells. We repeated this experiment using purified HS-5 exosomes and again observed that a fraction of FGF2 was protected from digestion (Figure 2F). Addition of 0.1% Triton X-100 disrupted the lipid membrane and resulted in complete digestion of all protein. We found a similar digestion pattern with the exosomal transmembrane proteins CD9 and tsg101. We conclude that FGF2 is contained within ECVs and exosomes, however we cannot exclude that FGF2 may also be on the surface since partial FGF2 degradation was noted in intact HS-5 cells, ECVs and purified exosomes (Figure 2E–F).

HS-5 stromal cells overproduce ECVs

Since HS-5 CM is more protective than HS-27 CM (Manshouri et al., 2011; Traer et al., 2016; Weisberg et al., 2008), we suspected that ECVs may be more numerous in HS-5 CM. We chose several orthogonal methods to quantify vesicles in CM. First, we used nanoparticle tracking analysis to quantify and compare HS-5 and HS-27 ECVs (Figure 3A). In parallel, we employed the Virocyt Virus Counter, a flow cytometry-based technique developed to detect viruses, which also works well to quantify ECVs (Figure 3B). As a gold standard, negative stain transmission electron microscopy of purified HS-5 and HS-27 exosomes was also used to image and quantify exosomes by counting (30–100 nm diameter with cup-shape appearance characteristic for exosomes, Figure 3C). Finally, we used sucrose step-gradient fractionation of HS-5 and HS-27 ECVs to compare cell compartment and exosome-specific markers by immunoblot (Figure 3D). Exosomes layer primarily at the 15–30% sucrose interface as indicated by exosomal markers CD9 and tsg-101, and are increased in HS-5 cells compared to HS-27. Interestingly the receptor for FGF2, FGFR1, was also found to localize preferentially with HS-5 exosomes. With all methods, we consistently observed greater than two-fold excess of vesicles produced by HS-5 compared to HS-27 cells (see Figure 3—figure supplement 1 for additional data). Markers of nucleus (lamin A/C), endoplasmic reticulum (calreticulin) and mitochondria (Bcl-XL) were located in the 45–60% interface containing larger microvesicles and apoptotic bodies.

Figure 3 with 1 supplement see all
HS-5 cells secrete more exosomes than HS-27 cells.

Equal numbers of HS-5 and HS-27 cells were plated in RPMI with exosome-depleted FBS for 24 hr. The ECVs were pelleted by ultracentrifugation at 100,000 g for 2 hr at four degrees C and resuspended in PBS. ECVs were quantified by (A) Nanosight, a nanovesicle tracking analysis, (B) Virocyt Virus Counter, a proprietary flow cytometry using fluorescent dyes that stain both nucleic acid and protein, or (C) transmission electron microscopy. (D) HS-5 and HS-27 exosomes were collected by sucrose-step gradient and analyzed by transmission electron microscopy. Vesicles were quantified by counting in three 2 × 2 μm areas per sample. All experiments were done in triplicate, error bars represent standard deviation, p values are indicated by *<0.05, **<0.005. HS-5 and HS-27 ECVs (P100) were obtained by ultracentrifugation (starting material, or SM), and the exosome fraction was further purified by a sucrose step-gradient. Sucrose layer interfaces (0–7.5%, 7.5–15%, 15–30%, 30–45%, and 45%-pellet) were collected, lysed and analyzed by immunoblot. Blots were probed with antibodies against exosomal markers CD9 and tsg101; cell compartment markers: fibronectin, lamin A/C, BCL-XL; as well as FGFR1 and FGF2. The lanes with highest enrichment for CD9 and tsg-101, indicating exosomes, are marked below.

https://doi.org/10.7554/eLife.40033.008

FGF2-FGFR1 signaling promotes stromal growth and paracrine protection of leukemia

FGF2 is an autocrine signaling protein for stroma, but recombinant FGF2 also mediates paracrine protection of leukemia cells (Traer et al., 2016; Traer et al., 2014). Thus there are two potential mechanisms by which FGFR inhibition can attenuate protection of leukemia cells in the marrow microenvironment: (1) FGFR inhibitors block FGF2-mediated paracrine protection at the leukemia cells; and/or (2) FGFR inhibitors interrupt stromal FGF2-FGFR1 autocrine signaling to reduce secretion of protective FGF2-containing exosomes. To compare the relative effect of FGFR inhibition on autocrine and paracrine signaling, HS-5 cells were pre-treated with the FGFR inhibitor PD173074 (Mohammadi et al., 1998; Trudel et al., 2004) for one week prior to collection of CM. CM collected from HS-5 cells pre-treated with PD173074 was significantly less protective than CM from an equal number of untreated HS-5 cells (Figure 4A), providing evidence that interruption of FGF2-FGFR1 signaling affects subsequent protection of leukemia cells. In contrast, addition of PD173074 to untreated HS-5 CM only modestly attenuated protection of MOLM14 cells. We found similar results with K562 cells exposed to imatinib (Figure 4—figure supplement 1). Purified ECVs from HS-5 CM, which are enriched in FGF2, were more sensitive to FGFR inhibition (Figure 1—figure supplement 1), however pre-treatment of HS-5 cells with PD173074 still had the greatest absolute reduction in protection. These results indicate that FGFR inhibitors overcome protection of leukemia cells primarily by directly altering secretion of FGF2-expressing stromal cells, making them significantly less protective.

Figure 4 with 3 supplements see all
FGF2 is an autocrine growth factor in bone marrow stromal cells, and FGFR inhibition attenuates growth.

(A) HS-5 cells were cultured in media ± 250 nM PD173074 for one week and then equal numbers of cells were replated for comparison. After adhesion, the cells cultured in PD173074 were washed and fresh media added to collect CM. MOLM14 cells were resuspended in media, untreated HS-5 CM, and PD pre-treated HS-5 CM and treated with ± 10 nM AC220 and ± 250 nM PD173074. Viability was measured by MTS assay after 72 hr and values were normalized to the relevant UT control. Error bars represent standard deviation, p values are indicated by *<0.05, **<0.005, and ***=0.0007. (B) HS-5 and HS-27 cells were plated in triplicate on 96 well plates in a gradient of FGFR inhibitor PD173074. Proliferation was measured using MTS reagent after 72 hr. Error bars indicate standard deviation. (C) HS-5 and HS-27 cells were incubated media ± 250 nM PD173074 (PD). The number of viable cells was measured with Guava ViaCount every 3 days over a 15 day period. Fresh media and PD173074 was added every 3 days. (D) HS-5 and HS-27 cells were incubated in media ± 1 µM PD173074 for 1 week. Brightfield microscopy images were obtained using a 10X objective. (E) HS-5 cells were incubated in 4-well glass chamber slides in media ± 250 nM PD173074 (PD). Cells were stained with lipophilic tracer DiI for 24 hr, fixed, then nuclei stained with DAPI. Immunofluorescent images were analyzed with CellProfiler software to determine cell size (μm [Weisberg et al., 2009]) and number of cells for each size range was binned and graphically displayed. PD173074 had no effect on HS-27 growth, morphology or size, consistent with an on-target FGFR effect. (F) Ex vivo cultured primary bone marrow stromal cells from a series of leukemia patients (n = 42) were lysed for RNA extraction and cDNA synthesis. Taqman qPCR analysis was performed using FGFR1, FGFR2, FGFR3, FGFR4, and FGF2 Taqman primer assays and expression plotted (n = 42 for each except FGFR4 which is n = 41 due to failed PCR for one sample). (G) FGFR1 and FGF2 qPCR values (2^-ΔCT) were plotted against each other. There were 9 AML patients with FLT3 ITD (most newly diagnosed) and these patients are indicated with red dots. Linear regression produced a line fit with r2 = 0.5683 and slope significantly non-zero with p<0.0001.

https://doi.org/10.7554/eLife.40033.010

To further evaluate the effects of FGFR inhibition in stromal cells, HS-5 cells were evaluated for viability, morphology, and growth using HS-27 cells as comparison (low FGF2). HS-5 or HS-27 cells had little reduction in cell viability after 72 hr treatment with PD173074 (Figure 4B), however HS-5 growth slowed dramatically over 15 days (Figure 4C). HS-5 cells exposed to PD173074 changed morphology and became less refractile, larger, and more adherent (Figure 4D). Cell size was quantified using CellProfiler software and PD173074 significantly increased HS-5 cell size (Figure 4E).

To evaluate FGF2 and FGFR1 expression in primary leukemia stroma, bone marrow aspirates from a series of leukemia patients were cultured ex vivo and FGF2 and FGFR1-4 expression quantified by RT-PCR (Figure 4F). FGFR1 and FGF2 transcripts were the most highly expressed in primary stroma, and there was a strong positive correlation between FGFR1 and FGF2 expression (Figure 4G, r2 = 0.5683 and p<0.0001 on nonparametric correlation). This indicates that FGF2 and FGFR1 expression are coordinately regulated in primary marrow stromal cells consistent with activation of an FGF2-FGFR1 autocrine loop. There were nine stromal cultures from AML patients with FLT3 ITD (indicated with red dots), but most of them were newly diagnosed, and based upon our previous data we would not expect increased expression of FGF216. Similar to our observations in cell lines described above, we also detected FGFR1 and FGF2 in ECVs derived from primary marrow stromal cultures (Figure 4—figure supplement 2). However, primary marrow stromal cells grow slowly and produce smaller amounts of ECVs, so we were unable to evaluate the effect of FGFR inhibitors on cell morphology, growth, and ECV production with primary marrow stromal cells. Additional characterization of primary stromal cultures is contained in Figure 4—figure supplement 3.

FGFR inhibition decreases stromal cell production of exosomes

Since FGFR inhibition attenuates HS-5 growth and morphology, we hypothesized that it might also reduce secretion of ECVs. HS-5 cells exposed to graded concentrations of PD173074 and BGJ-398 had a dose-dependent decrease in ECVs measured by Virocyt Virus Counter (Figure 5A,B). Notably, there was a significant decrease in vesicle number as early as 6 hr after drug exposure (Figure 5—figure supplement 1), suggesting that FGFR inhibition directly affects vesicle production or release. ECVs were also collected from HS-5 and HS-27 cells exposed to PD173074 and analyzed by immunoblot. PD173074 reduced the exosome markers tsg101 and CD9 (and FGF2) but had no effect on ECV production from HS-27 cells (Figure 5C, similar results with BGJ398 shown in Figure 5—figure supplement 1). Scanning electron microscopy of HS-5 cells revealed abundant budding membrane, whereas the surface of PD173074-exposed cells was smoother, implicating a change in membrane dynamics (Figure 5—figure supplement 2). To evaluate exosome secretion specifically, sucrose step-gradient fractionation was performed on ECVs from untreated and PD173074 treated HS-5 cells. PD173074 reduced exosomal markers CD9, tsg101, and FGF2 in the expected 15–30% interface fraction (Figure 5D).

Figure 5 with 2 supplements see all
FGFR inhibition decreases exosome production in FGF2-expressing stroma.

HS-5 cells were exposed to a gradient of the FGFR inhibitors (A) PD173074 and (B) BGJ-398 for 48 hr prior to collecting CM. ECVs were pelleted by ultracentrifugation at 100,000 g and quantified by Virocyt Virus Counter. Error bars indicate standard deviation and p values are indicated by *<0.05. (C) HS-5 and HS-27 cells were incubated in media ± 1 µM PD173074 for 72 hr prior to collecting ECVs. ECVs were analyzed by immunoblot for FGF2. The exosome markers CD9 and tsg101 are also shown. (D) HS-5 cells were plated in media ± 1 µM PD173074 for 72 hr. P100 fractions were obtained by ultracentrifugation, and further fractionated on a sucrose step-gradient. The interfaces (0–7.5%, 7.5–15%, 15–30%, 30–45%, and 45%-pellet) were collected, lysed and processed by immunoblot with antibodies against the exosomal markers CD9 and tsg101 as well as FGFR1 and FGF2.

https://doi.org/10.7554/eLife.40033.014

Genetic knock-down of FGFR1 or FGF2 attenuates exosome production

To confirm that decreased exosome secretion is specific for FGFR1 inhibition, HS-5 cells were stably transfected with either a GFP-expressing lentivirus control vector (GIPZ), or doxycycline-induced shRNA targeting FGFR1. FGFR1 silencing led to a significant reduction in ECVs (Figure 6C). Similar results were obtained with siRNA targeting FGFR1 (Figure 6—figure supplement 1). siRNA and shRNA constructs targeting FGF2 did not achieve reliable silencing of FGF2. HS-5 CRISPR/Cas9 knockout of FGFR1 and FGF2 in HS-5 cells were generated, however genetic silencing prevented continued growth. Multiple attempts to make stable deleted cell lines were unsuccessful, likely due to the importance of FGF2-FGFR1 signaling for HS-5 self-renewal and growth (Bianchi et al., 2003; Coutu et al., 2011; Zhou et al., 1998). That being said, ECVs collected shortly after CRISPR/CAS9 treatment, which resulted in partial silencing of FGF2 or FGFR1, both demonstrated decreased ECVs by immunoblot and reduced protection of MOLM14 cells (Figure 6—figure supplement 2). To test the role of FGF2 in ECV production in primary cells, equal numbers of murine stromal cells from Fgf2 +/+ and -/- mice (Fgf2tm1Doe [Zhou et al., 1998]) were treated with PD173074 and ECVs quantified by Virocyt (Figure 6D). Fgf2 +/+ stromal cells secreted significantly more ECVs than -/-, and PD173074 only reduced ECV secretion in +/+ stroma. ECVs from Fgf2 +/+ and -/- mice were also analyzed by immunoblot with similar reduction in ECV proteins from Fgf2 -/- stroma (Figure 6E).

Figure 6 with 2 supplements see all
Genetic silencing of FGFR1 or deletion of FGF2 attenuates exosome secretion.

A doxycycline-inducible lentiviral shRNA targeting FGFR1 was used to create a stable HS-5 cell line. The cells were then treated with doxycycline to induce FGFR1 silencing and compared to a GIPZ lentiviral control. (A) Silencing of FGFR1 expression is shown by immunoblot of cell lysates. ECVs from doxycycline-treated cells were analyzed by (B) immunoblot or (C) Virocyt Virus Counter. *p<0.05. (D) Bone marrow was isolated from Fgf2 +/+ and -/- mice and cultured ex vivo to grow adherent marrow stroma. Equal numbers of cells were then plated, CM collected for 72 hr, and then ultracentrifuged to collect ECVs. The ECVs were quantified by Virocyt. *p<0.05. (E) Equal number of cultured marrow cells from Fgf2 +/+ and -/- mice were plated and then ECVs collected by ultracentrifugation and analyzed by immunoblot.

https://doi.org/10.7554/eLife.40033.017

Fgf2 -/- stroma produces fewer exosomes and is less protective of BCR-ABL leukemia

To test the role of stromal Fgf2 in an in vivo leukemia model, bone marrow from Fgf2 +/+ mice was retrovirally transfected with BCR-ABL containing GFP as a marker (Traer et al., 2012) and used to transplant lethally irradiated FGF2 +/+ and -/- mice. This induces a very aggressive disease in mice that is more akin to AML than CML, and TKIs are only effective for a limited duration. Mice were treated with the ABL inhibitor nilotinib 75 mg/kg/day by oral gavage starting on day 14 post-transplant. Mice that were found to have aplastic marrow (unsuccessful transplantation) were excluded from analysis since their death was not related to leukemia (four mice in the Fgf2 +/+ untreated group, one mouse in the Fgf2 -/- untreated group, two mice in the Fgf2 +/+ nilotinib group, and two mice in the Fgf2 -/- nilotinib group). The survival curves of the remaining mice are shown in Figure 7A. The cohorts of untreated Fgf2 +/+ and -/- both died rapidly from disease, as expected. Nilotinib significantly increased survival of Fgf2 +/+ and -/- mice compared to untreated mice, but the survival of the nilotinib-treated Fgf2 -/- was also significantly longer than their Fgf2 +/+ counterparts. To ensure equal engraftment of disease in both backgrounds, the blood and bone marrow was analyzed for GFP and found to be similar in both Fgf2 +/+ and -/- mice at time of death (Figure 7B and Figure 7—figure supplement 1), suggesting that nilotinib was more effective at attenuating disease progression of BCR-ABL leukemia cells in an Fgf2 -/- microenvironment. To directly evaluate the protective effect of ECVs on leukemia progenitor cells, ECVs were isolated from equal numbers of +/+ and -/- primary marrow stromal cells cultured with and without PD173074 treatment. Then, bone marrow from +/+ mice was retrovirally transfected with BCR-ABL and incubated with the ECVs overnight. The cells were washed, plated in methylcellulose with and without imatinib, and colonies counted after 8 days. Imatinib significantly reduced colony formation without ECVs, but ECVs from +/+ stroma almost completely reversed the inhibitory effects of imatinib (Figure 7C). ECVs from PD173074-treated +/+ stroma or -/- stroma were not as protective, suggesting that Fgf2 +/+ stroma more effectively protects BCR-ABL leukemia cells from the effects of kinase inhibition through secretion of protective exosomes. We confirmed the presence of Fgf2 in microvesicles isolated from cultured Fgf2 +/+ mouse stroma (Figure 6). To confirm that ECVs can be endocytosed by primary cells, lineage-negative hematopoietic progenitor cells were isolated from Fgf2 +/+ mice and stained with a green lipophilic tracer (DiO) and incubated with ECVs from Fgf2 +/+ or Fgf2 -/- stromal cells stained with a red lipophilic tracer (DiI). Confocal microscopy confirmed internalization of fluorescently labeled primary stromal ECVs by murine progenitor cells (Figure 7D and Figure 7—video 1 and 2).

Figure 7 with 3 supplements see all
Fgf2 -/- mice survive significantly longer with TKI therapy in a murine BCR-ABL leukemia model.

Fgf2 +/+ bone marrow was removed from donor mice and spinoculated with pMIG BCR-ABL retrovirus containing an IRES-GFP marker. The transfected bone marrow was then transplanted into lethally irradiated Fgf2 +/+ or -/- recipients. Mice were treated with 75 mg/kg/day nilotinib by oral gavage starting on day 11 of transplant. (A) Survival curves of untreated and nilotinib-treated Fgf2 +/+ and -/- mice. (B) GFP in peripheral blood was evaluated weekly and at time of euthanasia to quantify disease burden. The average GFP (percent of nucleated cells) is shown and did not differ significantly between groups indicating that all animals developed similar disease burden. Error bars indicate standard deviation. (C) Bone marrow cells from Fgf2 +/+ mice were spinoculated with pMIG BCR-ABL retrovirus containing GFP-IRES. The cells were then incubated with ECVs obtained from Fgf2 +/+ and -/- primary stroma cultured alone or with 500 nM PD173074. The next day the incubated cells were washed three times to remove cytokines and exosomes and plated in cytokine-free methylcellulose ± imatinib. After 8 days, colonies were counted and normalized to untreated condition. Graph shown on right. Error bars indicate standard error of the mean. *p<0.05 and **p<0.005. (D) Lineage-negative bone marrow cells were isolated from Fgf2 +/+ mice and cells were stained with DiO (green) tracer, washed, and immobilized on Poly-D-lysine coated chamber slides. ECVs from bone marrow stroma of Fgf2 +/+ or -/- mice were stained with DiI (red) tracer and added to the cells for a 24 hr incubation. Slides were stained with DAPI (blue) and imaged by confocal fluorescent microscopy. Movie of the z-stack images are included as Figure 7—video 1 and 2. (E) Model of bone marrow stromal FGF2 autocrine signaling and paracrine protection of leukemia cells by FGF2-containing exosomes.

https://doi.org/10.7554/eLife.40033.020

Discussion

The normal hematopoietic microenvironment is altered by leukemia, and can protect leukemia cells from the effects of both chemotherapy and targeted kinase inhibitors (Yang et al., 2014; Parmar et al., 2011; Manshouri et al., 2011; Traer et al., 2014; Traer et al., 2016). Until recently, stromal protection of leukemia cells was thought to be largely mediated by secreted cytokines or through direct contact (review [Meads et al., 2009]). Here, we show that exosomes from bone marrow stromal cells are transferred to leukemia cells, and protect them from kinase inhibitors. Exosomes have previously been identified as important mediators of malignancy, including recent reports of leukemia exosomes modulating marrow stroma (Paggetti et al., 2015Huan et al., 2015; Peinado et al., 2012; Filipazzi et al., 2012). We found that the reciprocal transfer also occurs, and that marrow stromal exosomes efficiently protect leukemia cells from targeted kinase inhibitors. Along with recent reports that entire mitochondria are transferred between stromal cells and leukemia cells during therapy (Marlein et al., 2017; Moschoi et al., 2016), our data adds to an increasingly complex and intimate relationship between marrow stromal cells and leukemia cells. Indeed, it is almost hard to imagine the leukemia cell in the niche as a separate entity given the direct exchange of organelles, ECVs, cell-cell signaling, and secreted cytokine signaling between stromal and leukemia cells. A better understanding of this relationship is important to develop better ways to eradicate leukemia cells and cure more patients.

Isolated reports have previously suggested that FGF2 is contained in ECVs (Proia et al., 2008; Choi et al., 2016), but FGF2 has also been reported to self-assemble into a pore-like structure on the cell membrane and mediate its own translocation with the help of extracellular heparan sulfate (Steringer et al., 2015; Schäfer et al., 2004). Compared to other soluble secreted cytokines, FGF2 was uniquely enriched in ECVs and exosomes (Figure 2), suggesting that secretion in ECVs is the primary mechanism of FGF2 paracrine signaling from marrow stromal cells. Since FGFR1 is also found on exosomes (Figure 3D), the FGF2-FGFR1 interaction on exosomes may play a direct role in loading FGF2 in exosomes and/or regulate secretion. FGFR inhibitors also increase the amount of FGFR1 protein in stromal cells as measured by immunoblot, consistent with a role in receptor cycling and/or reduced secretion in exosomes.

Similar to our observations, epidermal growth factor receptor has been shown to be secreted on ECVs, and secretion is increased after ligand stimulation (Sanderson et al., 2008; Perez-Torres et al., 2008). Likewise, overexpression of oncogenic HER2 in breast cancer cell lines resulted in qualitative differences in microvesicle content (Amorim et al., 2014), suggesting a role for activated receptor tyrosine kinases in exosome production and secretion. Receptor-mediated endocytosis is the first step of exosome biogenesis (Théry et al., 2002), suggesting that inhibitors of receptor tyrosine kinases may act at this step. How FGFR1 is positioned in the exosome membrane (inside or out), how FGF2 binds FGFR1 in exosomes, and how exosomal FGF2 activates FGFR1 in leukemia cells, are areas of active investigation.

FGF2 has been previously implicated in hematologic malignancy progression and development of resistance (Sato et al., 2002; Chesi et al., 2001). Elevated levels of FGF2 have previously been measured in the serum of CML and AML patients (Aguayo et al., 2000; Aguayo et al., 2002), as well as in the bone marrow of AML patients, where it was reported to function as an autocrine promotor of proliferation (Bieker et al., 2003). We found that FGF2 expression was increased in CML and AML stroma during the development of resistance to kinase inhibitors, indicating that FGF2 expression is a regulated autocrine growth factor for stroma (Traer et al., 2016). This is consistent with the role of FGF2-FGFR1 autocrine expansion of stroma in stress-induced hematopoiesis (Itkin et al., 2012; Zhao et al., 2012) and suggests that leukemia cells are able to hijack the FGF2 stress response for survival. The regulation of FGF2-FGFR1 signaling is also supported by the positive correlation in expression of both FGF2 and FGFR1 in a subset of primary AML marrow samples (Figure 4G), indicating that this pathway can be selectively activated. FGFR inhibitors not only inhibit autocrine growth of stroma, but reduce exosome secretion and significantly alter the protective ability of stromal cells (Figures 4A and 7). Since exosomes contain a complex mixture of proteins, cytokines, lipids and microRNAs (all of which potentially contribute to leukemia cell protection), inhibiting secretion of exosomes is a promising approach to blunting this complex mechanism of resistance.

In summary, FGF2 is a regulated autocrine growth factor for marrow stroma that is important in reprogramming the marrow stroma during development of resistance to TKIs. FGF2-FGFR1 activation in marrow stroma leads to increased secretion of exosomes, which are protective of leukemia cells in both in vitro and in vivo models. Given the inevitable development of clinical resistance to TKIs (FLT3 ITD AML in particular), addition of FGFR inhibitors to directly modulate the leukemia niche is a promising approach to improve the durability of response.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional information
Gene
(homo sapiens)
FGF2NA
Gene
(mus musculus)
Fgf2NA
Gene
(homo sapiens)
FGFR1NA
Gene
(mus musculus)
Fgfr1NA
Strain, strain
background
(mus musculus)
Fgf2tm1Doe/J Fgf2 +/+ and -/- miceJackson LaboratoryRRID:MGI:2679603
Genetic reagent
(homo sapiens)
FGF2Thermo Fisher ScientificshRNA in
TRIPZ lentiviral vector
Genetic reagent
(homo sapiens)
FGFR1Thermo Fisher ScientificshRNA in
TRIPZ lentiviral vector
Genetic reagent
(homo sapiens)
AddGeneGeCKO lentiCRISPRv2
hSpCas9 and
guide RNA
Genetic reagent
(homo sapiens)
FGF2-1GenScriptCRISPR/Cas nine guide
RNA design
Genetic reagent
(homo sapiens)
FGF2-2GenScriptCRISPR/Cas nine guide
RNA design
Genetic reagent
(homo sapiens)
FGFR1-1GenScriptCRISPR/Cas nine guide
RNA design
Genetic reagent
(homo sapiens)
FGFR1-2GenScriptCRISPR/Cas nine guide
RNA design
Genetic reagent
(mus musculus)
pMIG with
BCR-ABL and
GFP
murine retrovirus
Cell line
(homo sapiens)
MOLM14Dr. Yoshinobu MatsuoRRID:CVCL_7916
Cell line
(homo sapiens)
K562American Type
Culture Collection
RRID:CVCL_0004
Cell line
(homo sapiens)
HS-5Dr. Beverly Torok-StorbRRID:CVCL_3720
Cell line
 (homo sapiens)
HS-27Dr. Beverly Torok-StorbRRID:CVCL_0335
AntibodyMouse monoclonal
anti-FGFR1
Cell Signaling9740Dilution 1:1000
AntibodyRabbit polyclonal anti-FGF2Santa CruzSc-79Dilution
1:500
AntibodyRabbit monoclonal anti-CD63ABCAMab134045Dilution 1:1000
AntibodyRabbit polyclonal anti-CD9Santa CruzSc-9148Dilution
1:200
AntibodyMouse monoclonal anti-tsg-101Santa CruzSc-7964Dilution
1:200
AntibodyMouse monoclonal anti-actinMilliporeMAB1501Dilution 1:5000
Peptide,
recombinant
protein
FGF2 (human)Peprotech
Commercial
assay or kit
Thermo Scientific lentiviral transfection kit
Chemical
compound, drug
quizartinib (AC220)LC labs
Chemical
compound, drug
imatinibLC labs
Chemical
compound, drug
nilotinibSelleckChem
Chemical
 compound, drug
PD173074SelleckChem
Chemical
compound, drug
BGJ-398SelleckChem
Chemical
compound, drug
doxycyclineFisher
Software,
 algorithm
CellProfilerCell area

Cell lines

Request a detailed protocol

The human cell line MOLM14 was generously provided by Dr. Yoshinobu Matsuo (Fujisaki Cell Center, Hayashibara Biochemical Labs, Okayama, Japan). The human cell line K562 was obtained from the American Type Culture Collection (Manassas, VA, USA). The human stromal cell lines HS-5 and HS-27a were kindly provided by Dr. Beverly Torok-Storb (Fred Hutchinson Cancer Research Center, Seattle, WA). Cells were maintained in RPMI1640 media supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin/100 μg/ml streptomycin, 2 mM L-glutamine, and 0.25 μg/ml fungizone (referred to as R10) at 37°C in 5% CO2. Exosome-depleted FBS was pre-cleared by ultracentrifugation at 100,000 g for 2 hr at 4°C. Cell lines were validated by genetic and functional analysis based upon previous reported characteristics. Cell lines were tested monthly for mycoplasma infection and discarded if found to be infected.

ECV isolation

Request a detailed protocol

HS-5 cells grown to 90–100% confluence in 15 cm dishes were washed in 8 ml PBS, and incubated in 12 ml exosome-depleted R10 overnight. The media was collected, cleared of debris (2 × 2000 g spin, 10 min), and ultracentrifuged at 100,000 g for 2 hr at 4°C. The resulting supernatant (S100) was poured off, and 100 ul PBS was added to the ECV pellet (P100). This was shaken for 4 hr at 4°C at 2000 rpm. P100 was used fresh or stored at −80°C with 10% DMSO.

Sucrose density step-gradient

Request a detailed protocol

Layers of sucrose (60%, 45%, 30%, 15%, 7.5%, and 0%) were carefully pipetted into ultracentrifuge tubes. ECVs were added on top, and the tube ultracentrifuged at 100,000 g for 90 min at 4°C. The sucrose interfaces (45–60, 30–45, 15–30, 7.5–15, 0–7.5) were collected with a micropipette, washed in PBS, and pelleted at 100,000 g for 2 hr at 4°C.

ECV quantitation

Request a detailed protocol

 ECVs were quantified by Nanosight LM10 or by Virocyt Virus Counter 3100, following manufacturers’ protocols.

Inhibitors and cytokines

Request a detailed protocol

Quizartinib (AC220) was purchased from LC labs (Woburn, MA, USA). Nilotinib, PD173074 and BGJ-398 were purchased from SelleckChem (Houston, TX, USA). Imatinib was purchased from LC labs (Woburn, MA, USA). Recombinant FGF2 was purchased from Peprotech (Rocky Hill, NJ, USA).

Viability assays

Request a detailed protocol

Viability was assessed with MTS reagent, CellTiter 96 AQueous One Solution Proliferation Assay from Promega Corporation (Madison, WI, USA) or by Guava ViaCount flow cytometer assay (Millipore, Burlington, MA, USA).

Immunoblot analysis

Request a detailed protocol

Treated cells were washed in PBS before adding lysis buffer (Cell Signaling, Danvers, MA, USA) supplemented with Complete protease inhibitor (Roche, Indianapolis, IN, USA) and phosphatase inhibitor cocktail-2 (Sigma-Aldrich, St. Louis, MO, USA). Proteins were fractionated on 4–15% Tris-glycine polyacrylamide gels (Criterion gels, Bio-Rad), transferred to PVDF membranes, and probed with antibodies: FGFR1, fibronectin (Cell Signaling, Danvers, MA, USA); CD9, FGF2, calreticulin, tsg101 (Santa Cruz Biotechnology, Dallas, TX, USA), CD63 (Abcam, Boston, MA, USA), and actin (MAB1501, Millipore, Burlington, MA, USA).

Stromal cell cytokine ELISA

Request a detailed protocol

Stromal CM, S100 and ECVs were lysed with 0.1% NP-40 for 30 min, centrifuged 3,000 rpm for 10 mins, and 50 μl supernatant was incubated with the magnetic beads overnight and assayed as per manufacturer's instructions (Luminex Multiplex magnetic beads 30-plex Assay, Life Technologies).

Primary bone marrow stromal cultures

Request a detailed protocol

Bone marrow aspirates were obtained from AML patients after informed consent under the OHSU Institutional Research Board protocol IRB0004422, and were processed as previously described (Viola et al., 2016). After Ficoll, the red cell pellets were incubated with ACK for 30 min on ice to lyse red cells, and plated on 15 cm dishes in MEM-α supplemented with 20% fetal bovine serum (FBS), 100 U/ml penicillin/100 μg/ml streptomycin, 2 mM L-glutamine, and 0.25 μg/ml fungizone at 37°C in 5% CO2. After 48 hr, non-adherent cells were removed and new media was added. This step was repeated after an additional 24 hr. Cells were then incubated for 1–3 weeks with media changes every 7 days, until patchy proliferation became apparent. Cells were trypsinized and replated to facilitate homogenous growth. Cells were expanded over a maximum of 3 passages before use in experiments. Murine primary stroma was isolated from harvested femur marrow without ACK treatment and then cultured as above. Primary stromal samples were analyzed after >2 weeks growth.

Transmission electron microscopy

Request a detailed protocol

Stromal cell exosomes were isolated by sucrose step-gradient then washed in 0.22 μm filtered PBS. 10 μl was deposited onto glow discharged carbon formvar 400 mesh copper grids (Ted Pella 01822 F) for 3 min, rinsed 15 secs in water, wicked on Whatman filter paper 1, stained for 45 secs in filtered 1.33% (w/v) uranyl acetate, wicked and air dried. Samples were imaged at 120kV on a FEI Tecnai Spirit TEM system. Images were acquired as 2048 × 2048 pixel, 16-bit gray scale files using the FEI’s TEM Imaging and Analysis (TIA) interface on an Eagle 2K CCD multiscan camera.

Fluorescent confocal microscopy

Request a detailed protocol

MOLM14 and K562 cells were stained with DiO (Thermo Fisher) according to manufacturer’s protocol. HS-5 ECVs were stained with DiI (Thermo Fisher), washed with PBS, and collected by ultracentrifugation. For experiments using mouse bone marrow, cells were isolated from femurs and tibias, RBCs were lysed using ACK buffer (0.8% NH4Cl and 0.1 mMEDTA in KHCO3 buffer; pH 7.2–7.6), and lineage-negative cells were isolated by MACS cell separation with the human lineage cell depletion kit (Milteny Biotec). Cells were incubated with a cytokine mix (IL-3, IL-6, SCF) in addition to DiO. DiO-stained cells were combined with DiI-stained ECVs and incubated for 24 hr at 37°C. Cells were washed, placed on poly-D-lysine coated chamber slides, and DAPI-stained. Z-stack imaging was performed on an Olympus IX71 inverted microscope. Images were processes using the Fiji software package (Schindelin et al., 2012).

Proteinase K digestion

Request a detailed protocol

ECVs, or exosomes isolated by sucrose step-gradient, were resuspended in proteinase K buffer (Tris-HCl pH8, 10 mM CaCl2) and then incubated with 200 μg/ml proteinase K at room temp for 30 min. 5 μL 0.1 M PMSF and SDS loading buffer was added and samples were incubated at 98°C for 5 min to stop reaction prior to immunoblots.

Cell morphology analysis

Request a detailed protocol

HS-5 and HS-27 cells were grown to 90% confluence in 4-well chamber microscope slides in R10 ±250 nM PD173074. Cells were stained with lipophilic tracer DiI, washed, and stained with DAPI. Cells were imaged with Zeiss Axio Observer fluorescent microscope at 10X using AxioVision software. Images were uploaded into CellProfiler software and analyzed for cell size. Cell diameter was determined as diameter [μm]=pixels×0.394 μm2/pixel.

shRNA

Request a detailed protocol

TRIPZ inducible lentiviral FGF2 and FGFR1 shRNA were purchased from Thermo Fisher Scientific Dharmacon RNAi Technologies (Waltham, MA, USA), along with Dharmacon’s trans-lentiviral shRNA packaging kit with calcium phosphate transfection reagent and HEK293T cells. HS-5 and HS-27 cells were transfected with GIPZ control or FGFR1 TRIPZ, per manufacturer’s protocol. TurboRFP/shRNA expression was induced with 1 μg/ml doxycycline (Fisher) for 48 hr, cells were washed in PBS, and then media replaced with exosome-depleted R10 +1 μg/ml doxycycline. Cells and CM were collected after 72 hr for analysis.

CRISPR/Cas9 targeted genome editing

Request a detailed protocol

 The vector GeCKO lentiCRISPRv2 was obtained from Addgene. This plasmid contains two expression cassettes, hSpCas9 and the chimeric guide RNA. Guide RNA sequences were obtained from GenScript (Sanjana et al., 2014), and oligos with 5’ overhang for cloning into lentiCRISPRv2 were manufactured by Fisher Scientific. The vector was digested with BsmBI and dephosphorylated, the plasmid was gel-purified, and oligonucleotides were ligated after annealing and phosphorylation. Plasmid was amplified in Stbl3 bacteria, purified, and lentivirus was generated in HEK293T cells. Transduced HS-5 cells were selected in puromycin for 5 days, and cultured an additional 5 days before assessing knockout.

Murine BCR-ABL leukemia experiments

Request a detailed protocol

Animal studies were approved by the OHSU Institutional Animal Care and Use Committee. Fgf2tm1Doe/J were purchased from Jackson Laboratory to breed homozygous +/+ and -/- littermates. Bone marrow from 5-FU treated Fgf2 +/+ donors was spinoculated with pMIG containing BCR-ABL and IRES-GFP reporter as previously described (Traer et al., 2012; Agarwal et al., 2008) and 2 × 106 cells were retro-orbitally injected into lethally irradiated (2 × 450 cGy administered 4 hr apart) Fgf2 +/+ and -/- recipients. 75 mg/kg/day nilotinib was administered by oral gavage and mice were monitored weekly with cell blood counts and FACS analysis of GFP in peripheral blood. Diseased mice were subjected to detailed histopathologic analysis. For colony assays, ECVs were isolated (as above) from equal numbers of Fgf2 +/+ and -/- primary stromal cells cultured on 10 cm plates for 3 days, with and without 500 nM PD173074 (3 day pre-treatment and 3 days during ECV collection). Bone marrow from FGF2 +/+ mice was spinoculated with pMIG containing BCR-ABL and IRES-GFP reporter as above, incubated with ECVs overnight and washed 3x the next day. 3% of cells were GFP positive by FACS, and 4 × 103 cells were then plated in 1 ml of MethoCult M3234 Methylcellulose Medium for Mouse Cells without cytokines (Stemcell Technologies) in triplicate. Mouse bone marrow colonies larger than 50 cells were counted after 8 days.

Statistical methods

Request a detailed protocol

Graphical and statistical data were generated with Microsoft Excel or GraphPad Prism (GraphPad Software, La Jolla, CA, USA). P value < 0.05 was considered significant.

Conflict of interest disclosure

Request a detailed protocol

The authors declare no competing interests. Dr. Druker is currently principal investigator or co-investigator on Novartis clinical trials. His institution, OHSU, has contracts with these companies to pay for patient costs, nurse and data manager salaries, and institutional overhead. He does not derive salary, nor does his lab receive funds from these contracts.

References

  1. 1
  2. 2
    Angiogenesis in acute and chronic leukemias and myelodysplastic syndromes
    1. A Aguayo
    2. H Kantarjian
    3. T Manshouri
    4. C Gidel
    5. E Estey
    6. D Thomas
    7. C Koller
    8. Z Estrov
    9. S O'Brien
    10. M Keating
    11. E Freireich
    12. M Albitar
    (2000)
    Blood 96:2240–2245.
  3. 3
  4. 4
  5. 5
    Modulation of megakaryocytopoiesis by human basic fibroblast growth factor
    1. H Avraham
    2. N Banu
    3. DT Scadden
    4. J Abraham
    5. JE Groopman
    (1994)
    Blood 83:2126–2132.
  6. 6
  7. 7
    Basic fibroblast growth factor mediates its effects on committed myeloid progenitors by direct action and has no effect on hematopoietic stem cells
    1. AC Berardi
    2. A Wang
    3. J Abraham
    4. DT Scadden
    (1995)
    Blood 86:2123–2129.
  8. 8
  9. 9
    Overexpression of basic fibroblast growth factor and autocrine stimulation in acute myeloid leukemia
    1. R Bieker
    2. T Padró
    3. J Kramer
    4. M Steins
    5. T Kessler
    6. S Retzlaff
    7. F Herrera
    8. J Kienast
    9. WE Berdel
    10. RM Mesters
    (2003)
    Cancer Research 63:7241–7246.
  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
  30. 30
  31. 31
    Functionally distinct human marrow stromal cell lines immortalized by transduction with the human papilloma virus E6/E7 genes
    1. BA Roecklein
    2. B Torok-Storb
    (1995)
    Blood 85:997–1005.
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54

Decision letter

  1. Martin McMahon
    Reviewing Editor; University of Utah Medical School, United States
  2. Jeffrey Settleman
    Senior Editor; Calico Life Sciences, United States
  3. Thomas O'Hare
    Reviewer; University of Utah School of Medicine, 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 all the issues have been addressed.]

Thank you for submitting your article "FGF2-FGFR1 signaling regulates release of leukemia-protective exosomes from bone marrow stromal cells" for consideration by eLife. Your article has been reviewed by two 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. The following individual involved in review of your submission has agreed to reveal their identity: Thomas O'Hare (Reviewer #2).

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:

This manuscript addresses the mechanism(s) by which protective signaling from the leukemia microenvironment may promote leukemia cell persistence, the development of drug resistance, and ultimately disease relapse. The authors claim that fibroblast growth factor 2 (FGF2) is released in exosomes from bone marrow stromal cells and that the subsequent endocytic entry of FGF2 into leukemia cells protects the leukemia cells from oncoprotein kinase targeted inhibitors. The authors claim that expression of FGF2, and its receptor FGFR1, is increased in a subset of stromal cell lines and primary AML stroma. Moreover, they claim that increased FGF2/FGFR1 signaling is associated with increased exosome secretion such that pharmacologic or genetic inhibition of FGFR1 signaling inhibition interrupts stromal autocrine growth and decreases secretion of FGF2-containing exosomes resulting in decreased protection of leukemia cells. Support for this hypothesis comes from the observation that, when BCR-ABL-driven leukemia cells are transplanted into FGF2-deficient mice, they display enhanced therapeutic benefit compared to FGF2-proficient mice. Hence, the authors claim that inhibition of the FGF2->FGFR1 signaling axis may be a strategy to prevent or overcome the development of TKI resistance in leukemia. Overall, the findings of this study are supported by strong, well-controlled experimental data and the manuscript is clearly written. However, both reviewers have both major and minor concerns, which if addressed by the authors, will substantially improve the quality of the manuscript.

Separate reviews (please respond to each point):

Reviewer #1:

This manuscript address the mechanism(s) by which protective signaling from the leukemia microenvironment may promote leukemia cell persistence, the development of drug resistance, and ultimately disease relapse. The authors claim that fibroblast growth factor 2 (FGF2) is released in exosomes from bone marrow stromal cells and that the subsequent endocytic entry of FGF2 into leukemia cells protects the leukemia cells from oncoprotein kinase targeted inhibitors. The authors claim that expression of FGF2 and its receptor FGFR1, are increased in a subset of stromal cell lines and primary AML stroma. Moreover, they claim that increased FGF2/FGFR1 signaling is associated with increased exosome secretion such that pharmacologic or genetic inhibition of FGFR1 signaling inhibition interrupts stromal autocrine growth and decreases secretion of FGF2-containing exosomes resulting in decreased protection of leukemia cells. Support for this hypothesis comes from the observation that, when BCR-ABL driven leukemia cells are transplanted into FGF2 deficient mice, they display enhanced therapeutic benefit compared to FGF2 proficient mice. Hence, the authors claim that inhibition of the FGF2->FGFR1 signaling axis may be a strategy to prevent or overcome the development of TKI resistance in leukemia.

The authors conclude from the data in Figure 1C that ECVs produced by HS-5 cells are internalized by endocytosis. What evidence do the authors have the ECVs are actually entering cells by endocytosis and, if entry is by endocytosis, is it clathrin-mediated endocytosis, caveolae, micropinocytosis or phagocytosis?

The data in Figures2E/F on protection of exosomal FGF2 from proteinase K digestion are quite equivocal in that the partial protection observed is not entirely consistent with the authors' interpretations. Indeed, the protection described in Figure 2F is so marginal as to make this reviewer wonder how reproducible the data are and whether, or not, a modest increase in incubation time with proteinase K would lead to complete proteolysis of FGF2. Indeed, the authors' conclusions that, "….. FGF2 is contained within exosomes but also likely associated with proteins on the surface of exosomes." is rather ambivalent.

In Figure 3, the authors convincingly demonstrate that HS-5 CM contains more exosomes that HS-27 CM. They also note that the receptor for FGF2, FGFR1, is also present in exosomes. One obvious question is whether FGFR1 is tyrosine phosphorylated on sites that are stimulated by FGF2, which would suggest that a portion of FGF2 and FGFR1 are in complex with one another.

In Figure 5, the authors show that TKIs of FGFR1 signaling reduce exosome release from HS-5 stromal cells. This data are convincing, but it is also rather phenomenological and lacks a mechanism by which FGFR1 might regulate the release of exosomes by bone marrow stromal cells. Also, the fact that this occurs in one cell line does not lend confidence that this is a generalizable conclusion.

In Figure 6, the use of Tet-regulated shRNA targeting shRNA is a good complement to the use of pharmacological TKIs of FGFR1 signaling. However, the effects of shFGFR1 on exosomal FGF2 should be assessed in Figure 6B – since the effects of shFGFR1 on tsg101 and CD9 are barely discernible given the poor quality of immunoblots presented in this figure. Furthermore, the authors need to better explain why their CRISPR/CAS9 efforts to silence the expression of either FGFR1 or FGF2 resulted in partial loss of FGFR1 or FGF2 expression. Moreover, whereas partial silencing of FGFR1 had almost no discernible effect on exosomal tsg101 or CD9, partial silencing of FGF2 had an appreciable effect on exosomal tsg101 and CD9. These data are somewhat discordant and are not adequately explained in the text.

The experiments presented in Figure 7 were nicely conceived and elegantly executed. This figure makes a case for an important role of stromal derived, exosome delivered FGF2 as a relevant mechanism of TKI resistance in leukemia.

Reviewer #2:

The findings of the study are supported by strong, well-controlled experimental data. The paper is clearly written. My major concerns are the heavy reliance on HS-5 as proxy for the leukemic microenvironment and that the story is almost entirely presented from the perspective of the stroma. Several mechanistic questions pertaining to the leukemic cells are asked below, with the goal of helping the reader to better understand how the authors are viewing and rationalizing these points.

1) A stated theme of the study is that the normal hematopoietic microenvironment is altered by leukemia. While this is undoubtedly correct, the choice of CM from a cell line derived from normal bone marrow would seem to limit applicability and relevance to the leukemic microenvironment.

2) Why doesn't the presence of an FGFR inhibitor (PD173074) in experiments in which CM is provided to leukemic cell lines dampen the effectiveness of HS-5 CM? The authors make a convincing case that FGF2-laden exosomes are taken up by the leukemic cell lines and that this provides a protective effect against TKIs. They also show that FGFR inhibition (or [partial] FGFR1 deletion) in HS-5 stromal cells decreases exosome production/secretion and probably alters the content of exosomes. However, the story from the point of view of the leukemic cell lines is largely unexplored and mechanistically vague. Does FGFR1 deletion in the leukemic cell lines change the effectiveness of HS-5 CM? Does PD173074 completely shut off FGFR1 signaling in the leukemic cell lines? Perhaps I am missing something, but I find it surprising that inhibition of FGFR1 (in the leukemic cells) does not at least partially reverse the protective effects of HS-5 CM. This is largely based on the authors' reasonable assertion that the most important exosome cargo is FGF2.

3) Several TKIs used in CML have target profiles that include FGFR, most notably the pan-FGFR TKI, ponatinib. Midostaurin, the currently approved frontline TKI for FLT3-mutated AML, inhibits FGFR1 and 2. Would patients treated with these TKIs derive additional benefit from collateral inhibition of FGFR1 in stromal cells (and possibly in leukemic cells)? I expected to see a comparison of imatinib and ponatinib in the study. I also thought this would be worth testing in the mouse model, as mentioned below.

4) Panels 4F,G show data for primary AML stromal cells (n = 42). The legend should specify how many AML patients are represented and how many data points are from each patient. Are some of these FLT3-ITD patients, and if so, can you indicate them as colored dots? This display item and the corresponding Materials and methods, Results and Discussion require more detail. This has potential to be an intriguing part of the story but the presentation is murky. Is it possible to include a comparison to HS-5 and HS-27 in these panels? Are the FGFR1 and FGF2 levels (and their correlations) substantially different from normal bone marrow? Should we expect them to be? Please consider strengthening this part of the manuscript. I could not glean much as presented.

Minor Comments:

1) Figure 1: In panel A, what is the result if the same experiment is carried out with ponatinib (10-25 nM)? dasatinib (10-25 nM)? In panel B, with midostaurin (1 µM)? This comment is somewhat redundant with major point (3) above; here, I am requesting that the data are included, at least in the supplement.

2) Why was 1 µM imatinib used? This concentration, at least at the 48 h timepoint, does not reduce proliferation (Figure 1A) to the extent that AC220 does in MOLM14 cells. Thus, it is difficult to determine the extent to which CM provides a protective effect. What happens if a higher concentration of imatinib is used, or if a more potent TKI is used at a concentration that matches the AC220 magnitude of reduced proliferation?

3) Regarding the FGFR inhibitors:

– Why two? BGJ-398 is described as 'selective' but I did not find the rationale for needing it in the study (which began with PD173074).

– What is the direct evidence/readout to confirm that these inhibitors are working (and to measure the extent of inhibition)?

4) Perhaps it's not a reasonable comparison, but it is surprising that shRNA-based elimination of FGFR1 does not reduce FGF2 levels in HS-5 cells (Figure 6A), but FGFR inhibition does (Figure 5C). Also, the HS-5 CRISPR FGFR1.1 and 1.2 lines show reduced FGF2. The HS-5 CRISPR FGF2.1 and 2.2 lines show enhanced FGFR1 but reduced FGFR1 in ECVs. Can the authors comment on / explain these points?

5) In Figure 6B, I would like to be able to evaluate whether reduced FGFR1 brings about a reduction in FGF2, but this data is not provided. In general, both FGF2 and FGFR1 should be shown in immunoblot stacks. Sometimes, it is one or the other.

6) Figure 7:

– why is nilotinib used in 7A? All other data are with imatinib.

– the main point of the paper seems to be that interrupting the FGFR1-FGF2 network in stromal cells would decrease the protective effects emanating from the microenvironment. To test this, shouldn't arms with "PD PD173074" alone, and with "nilotinib + PD173074" be included?

– why did you choose this model rather than a FLT3-ITD model? The jumping back and forth between CML and AML is distracting. Could the authors consider a focus on MOLM14 and FLT3-ITD driven AML in the main manuscript, with the CML work allocated to the supplement?

7) Please address the following clerical issues:

– abbreviations are defined several times (examples: TKI; conditioned media)

– conventions are not consistently followed (ml vs. mL; h vs. hrs vs. hours; spacing between value and unit [e.g. 10nM vs. 10 nM]; no spacing vs. spacing preceding a citation)

– Chemical formulas lack proper subscripting (possibly an inherent submission portal limitation)

– Please check supplier names (Genscript or GenScript? Is Jackson laboratories the correct name of this supplier?)

– References 14 and 30 are identical

– A few article titles in References section are title case (#4, 15, 16, for example)

– Reference 23 lacks page number, etc. The final manuscript is 2016, not 2015

– Reference 25 lacks journal name, etc.

– Figure 2 legend: NB40 or NP-40?

– Final paragraph of subsection “FGF2 is Contained in Stromal Cell ECVs and Exosomes”: "membrate"

– First paragraph of subsection “FGF2-Fgfr1 Signaling Promotes Stromal Growth and Paracrine Protection of Leukemia”: "withg"

– "proteinase K" vs. "Proteinase K"

– "PD173074" vs. "PD-173074"

– Second paragraph of the Discussion: a callout to Figure 3E is made. Should this be 3D (there is no panel 3E)?

Additional data files and statistical comments:

I asked the authors to consider including additional data based on experiments with tyrosine kinase inhibitors beyond imatinib and AC220. If they choose to do this, inclusion of the data files is warranted. I do not see the need for any other additional data files.

Statistical analysis and description of statistical methods is adequate and appropriate.

Experimental rigor is evident in the work presented in the manuscript.

https://doi.org/10.7554/eLife.40033.028

Author response

Reviewer #1:

[…] The authors conclude from the data in Figure 1C that ECVs produced by HS-5 cells are internalized by endocytosis. What evidence do the authors have the ECVs are actually entering cells by endocytosis and, if entry is by endocytosis, is it clathrin-mediated endocytosis, caveolae, micropinocytosis or phagocytosis?

The reviewer makes a good point, we do not present evidence that internalization is mediated by endocytosis. More detailed mechanistic studies of exosome transfer and uptake is in progress and beyond the scope of this manuscript. The text was modified as follows:

Analysis by confocal microscopy showed that ECVs are indeed internalized by leukemia cells, although the exact mechanism of internalization is still under investigation (Figure 1C and Supplemental movies).

The data in Figures2E/F on protection of exosomal FGF2 from proteinase K digestion are quite equivocal in that the partial protection observed is not entirely consistent with the authors' interpretations. Indeed, the protection described in Figure 2F is so marginal as to make this reviewer wonder how reproducible the data are and whether, or not, a modest increase in incubation time with proteinase K would lead to complete proteolysis of FGF2. Indeed, the authors' conclusions that, "….. FGF2 is contained within exosomes but also likely associated with proteins on the surface of exosomes." is rather ambivalent.

We repeated this experiment multiple times and it is technically challenging. Even the intact HS-5 cells in Figure 2E (mostly intracellular FGF2) have some degraded FGF2 after proteinase K treatment, likely from cell rupture. Isolated EVs and exosomes are even more fragile and we found that even integral membrane proteins tsg-101 and CD9 were partially degraded by proteinase K. The fact that we could detect intact FGF2 after proteinase K is strong evidence that some FGF2 (and perhaps most) is contained within EVs and exosomes. The text was modified to state this more explicitly.

We conclude that FGF2 is contained within EVs and exosomes, however we cannot exclude that FGF2 may also be present on the surface since partial FGF2 degradation was noted in intact HS-5 cells, EVs and purified exosomes (Figures 2E-F).

In Figure 3, the authors convincingly demonstrate that HS-5 CM contains more exosomes that HS-27 CM. They also note that the receptor for FGF2, FGFR1, is also present in exosomes. One obvious question is whether FGFR1 is tyrosine phosphorylated on sites that are stimulated by FGF2, which would suggest that a portion of FGF2 and FGFR1 are in complex with one another.

This is something we are quite interested in, however there are no reliable phospho-FGFR1 antibodies that we have found – and not for lack of trying! To address the issue of FGF2 bound to FGFR1 directly, we are generating FGF2 and FGFR1 expression constructs with bi-fluorescent complementation tags. This system produces fluorescence when FGF2 and FGFR1 are bound to each other, bringing the bi-fluorescent tags together to generate fluorescent signal. However, we do not have this data yet.

In Figure 5, the authors show that TKIs of FGFR1 signaling reduce exosome release from HS-5 stromal cells. This data are convincing, but it is also rather phenomenological and lacks a mechanism by which FGFR1 might regulate the release of exosomes by bone marrow stromal cells. Also, the fact that this occurs in one cell line does not lend confidence that this is a generalizable conclusion.

There are limited human marrow stromal cell lines available, but HS-5 and HS-27 are widely used as representative of distinct phenotypic subsets of marrow stroma and have clear differences in FGF2/FGFR1 expression. We were able to replicate reduction in exosomes with FGFR inhibition in primary stroma from FGF2 +/+ mice but not FGF2 -/- mice (Figures 6 and 7), which we feel is the best genetic control. We attempted to treat primary human stroma with FGFR inhibitors but these cultures are quite heterogeneous making comparisons difficult, and we were unable to obtain sufficient conditioned media from primary cultures for reliable exosome quantification after FGFR inhibition.

This FGFR1 downstream pathways that regulate exosome secretion is still under investigation in our lab. Some of our preliminary data is in Author response image 1 and suggests that FGFR1 regulates exosome biogenesis through phospho-lipase C and protein kinase C activation, but experiments are ongoing to prove this.

Ongoing model and mechanistic work on PLC. We reasoned that FGFR1 can either lead to increased exosome biosynthesis and/or release. One of the well-established FGFR signaling intermediaries is phospholipase C, a membrane-associated enzyme that cleaves phospholipids into diacyl glycerol (DAG) and inositol 1,4,5-triphosphoate (IP3). Secretory vesicle trafficking involves several steps that are controlled by DAG, including the fission of vesicles at the trans-Golgi network, the generation and maturation of multivesicular bodies, and the docking and fusion at the plasma membrane. Previous reports have shown that DAG kinase α (DGKα converts DAG to phosphatidic acid) expression decreases the production of FasL-containing exosomes by T lymphocytes, and that inhibition of DGKα enhances the production (Alonso et al. Cell Death Differ. 2011). DAG also activates protein kinase C (PKC), an enzyme which is recruited to the plasma membrane or to a number of intracellular compartments upon activation. PKC controls the endocytosis, trafficking and recycling of several receptor tyrosine kinases. This may be a relevant mechanism for the regulation of endocytosis and subsequent routing to multivesicular bodies. Although PKC has not previously been shown to phosphorylate FGFR1, FGFR1 does possess C-terminal serine/threonine phosphorylation sites that are known to be important for regulation of endocytosis(Nadratowska-Wesolowska et al. Oncogene. 2014).

We therefore sought to test the roles of PLC and PKC in exosome release in HS-5 cells. A rise in intracellular calcium is necessary for regulated exocytosis of secretory granules and exosomes. We therefore used the cell-permeable calcium ionophore ionomycin as a positive control for the regulation of exosome release. PD173074, which we confirmed as a negative regulator of exosome production in previous figures, served as a negative control. Phorbol 12-myristate 13-acetate (PMA) is a mimic of DAG and activates PKC. Over a timecourse of 72 hours, both PMA and ionomycin led to increased exosome production/release. PD173074 attenuated exosome production. Interestingly, this was not rescued by combined treatment with ionomycin, suggesting that FGFR inhibits early exosome biogenesis rather than release.

In Figure 6, the use of Tet-regulated shRNA targeting shRNA is a good complement to the use of pharmacological TKIs of FGFR1 signaling. However, the effects of shFGFR1 on exosomal FGF2 should be assessed in Figure 6B – since the effects of shFGFR1 on tsg101 and CD9 are barely discernible given the poor quality of immunoblots presented in this figure. Furthermore, the authors need to better explain why their CRISPR/CAS9 efforts to silence the expression of either FGFR1 or FGF2 resulted in partial loss of FGFR1 or FGF2 expression. Moreover, whereas partial silencing of FGFR1 had almost no discernible effect on exosomal tsg101 or CD9, partial silencing of FGF2 had an appreciable effect on exosomal tsg101 and CD9. These data are somewhat discordant and are not adequately explained in the text.

The FGF2 protein in ECVs after shRNA silencing of FGFR1 was assessed by immunoblot in this experiment, but unfortunately was uninterpretable. The amount of protein in ECVs is very low, and many of our immunoblots were at the limit of detection when performing small scale experiments (i.e. not using multiple 15 cm dishes of HS-5 cells to generate ECVs). The immunoblots for tsg-101 and CD9 in Figure 6 were darkened to allow better visualization of the bands.

The CRISPR/CAS9 experiments were complicated by the fact that complete silencing of either FGFR1 or FGF2 in HS-5 cells essentially halted cell replication and we were never able to continue to culture cells with stable FGF2 and FGFR genetic deletion. We attempted this experiment numerous times, but could only analyze the cells for a short period after initial silencing. For this reason, the CRISPR/CAS9 cells only have partial silencing of FGF2 and FGFR1 protein. We did notice that FGF2 silencing led to an increase in FGFR1 protein on the cells and more rapid decrease in ECVs. The increase in FGFR1 proteins is something we have observed with FGFR inhibitors previously (Traer et al. Blood 2014 and Traer et al. Cancer Research 2016) and based upon our previous data we suspect that removing FGF2 ligand leads to less activation and internalization of FGFR1. We suspect that silencing of FGFR1 takes longer to affect ECV production since there is still abundant FGF2 ligand present to drive activity of the remaining FGFR1. Given the limitations of our genetic silencing tools, we could not investigate this more rigorously. Given the limitations of using CRISPR/CAS9 for FGF2 and FGFR1 in HS-5 cells, this data was moved to the supplement, along with a more detailed explanation of the above issues.

In contrast to the CRISPR/CAS9 data, the primary stroma from FGF2 +/+ and -/- mice is a cleaner and more robust genetic control. Although the FGF2 -/- marrow stroma does not culture well in vitro, we could overcome this deficiency by simply using more mice. The immunoblots of ECVs from primary FGF2 +/+ and -/- stroma was moved to Figure 6 from the supplement.

We feel the presentation of the data this way is easier to understand, and the issues with the CRISPR/CAS9 system – although supportive overall of the model – does have technical limitations and is more appropriate for the Supplement.

The experiments presented in Figure 7 were nicely conceived and elegantly executed. This figure makes a case for an important role of stromal derived, exosome delivered FGF2 as a relevant mechanism of TKI resistance in leukemia.

We appreciate the positive feedback, thank you.

Reviewer #2:

The findings of the study are supported by strong, well-controlled experimental data. The paper is clearly written. My major concerns are the heavy reliance on HS-5 as proxy for the leukemic microenvironment and that the story is almost entirely presented from the perspective of the stroma. Several mechanistic questions pertaining to the leukemic cells are asked below, with the goal of helping the reader to better understand how the authors are viewing and rationalizing these points.

1) A stated theme of the study is that the normal hematopoietic microenvironment is altered by leukemia. While this is undoubtedly correct, the choice of CM from a cell line derived from normal bone marrow would seem to limit applicability and relevance to the leukemic microenvironment.

Although HS-5 is derived from normal stroma, it has long been recognized has having distinct effects on hematopoiesis compared to HS-27. We previously evaluated FGF2 in stromal cells from patients with resistant CML and AML and found significantly increased FGF2 expression by immunohistochemistry during development of resistance (Traer et al. Blood 2014, Traer et al. Cancer Research 2016). The dramatically different expression of FGF2 and FGFR1 in HS-5 compared to HS-27 make it a good model to study FGF2 secretion. Also, we were unable to generate sufficient numbers of ECVs from cultured primary stromal cells from leukemia patients to make this a feasible option using currently available methods for ECV quantification.

The data with primary murine stromal cells from FGF2 +/+ and -/- and the leukemia model in Figures 6 and 7 provides strong supportive data that marrow stromal FGF2 mediates resistance in the leukemia microenvironment specifically.

2) Why doesn't the presence of an FGFR inhibitor (PD173074) in experiments in which CM is provided to leukemic cell lines dampen the effectiveness of HS-5 CM? The authors make a convincing case that FGF2-laden exosomes are taken up by the leukemic cell lines and that this provides a protective effect against TKIs. They also show that FGFR inhibition (or [partial] FGFR1 deletion) in HS-5 stromal cells decreases exosome production/secretion and probably alters the content of exosomes. However, the story from the point of view of the leukemic cell lines is largely unexplored and mechanistically vague. Does FGFR1 deletion in the leukemic cell lines change the effectiveness of HS-5 CM? Does PD173074 completely shut off FGFR1 signaling in the leukemic cell lines? Perhaps I am missing something, but I find it surprising that inhibition of FGFR1 (in the leukemic cells) does not at least partially reverse the protective effects of HS-5 CM. This is largely based on the authors' reasonable assertion that the most important exosome cargo is FGF2.

Conditioned media from HS-5 cells contains a number of cytokines and other secreted factors in addition to FGF2-containing exosomes so the impact of adding an FGFR inhibitor such as PD173074 to CM is less pronounced in this setting (Figure 4A). If ECVs are isolated from HS-5 CM and then added to K562 or MOLM14 cells, then the inhibitory effect of PD173074 is more pronounced, although there is still some protection that is not blocked by FGFR inhibition. This is most likely due to protection from other components of exosomes/ECVs. To allow comparison with pure activation of FGFR, we did the same experiment with recombinant FGF2, in which case PD173074 completely blocks the protective effect. The data has been added to Figure 1—figure supplement 1.

The most impressive reduction in protection still comes from treating the HS-5 cells with PD173074 prior to collection of CM (Figure 4A). We found a similar reduction in protection with partial genetic silencing (Figure 6—figure supplement 2) and with genetic deletion of FGF2 from the leukemia microenvironment (Figure 7). This leads to reduced secretion of FGF2 in exosomes and likely has other effects on secretion of protective factors.

3) Several TKIs used in CML have target profiles that include FGFR, most notably the pan-FGFR TKI, ponatinib. Midostaurin, the currently approved frontline TKI for FLT3-mutated AML, inhibits FGFR1 and 2. Would patients treated with these TKIs derive additional benefit from collateral inhibition of FGFR1 in stromal cells (and possibly in leukemic cells)? I expected to see a comparison of imatinib and ponatinib in the study. I also thought this would be worth testing in the mouse model, as mentioned below.

We looked at ponatinib extensively in our previous CML publication (Traer et al. Blood 2014) and argue that additional FGFR inhibition contributes to the effectiveness of ponatinib in non-mutated, resistant CML patients. We have also found that midostaurin blocks many of the protective factors of stroma when MOLM14 cells are co-cultured with HS-5 cells (data not shown), similar to what we have previously published with quizartinib and PD173074 combination treatment (Traer et al. Cancer Research 2016). However, both ponatinib and midostaurin affect a number of additional pathways, especially at higher doses, so we chose to use specific inhibitors of FGFR in our experiments and the mouse model with deletion of FGF2 to get around the issue of off-target effects.

To further address this issue and minor point 1 below, we compared protection with recombinant FGF2, purified HS-5 ECVs, and CM with ECVs removed, using multiple ABL and FLT3 inhibitors. The data is now included in Figure 1—figure supplement 1. As predicted, higher doses of ponatinib and midostaurin are both able to overcome the protective effects of FGF2 and HS-5 ECVs at doses near the reported IC50 for FGFR (between 10-100 nM with ponatinib [Traer et al. Blood 2014] and 200nM midostaurin [Chen et al. PNAS 2014]).

4) Panels 4F,G show data for primary AML stromal cells (n = 42). The legend should specify how many AML patients are represented and how many data points are from each patient. Are some of these FLT3-ITD patients, and if so, can you indicate them as colored dots? This display item and the corresponding Materials and methods, Results and Discussion require more detail. This has potential to be an intriguing part of the story but the presentation is murky. Is it possible to include a comparison to HS-5 and HS-27 in these panels? Are the FGFR1 and FGF2 levels (and their correlations) substantially different from normal bone marrow? Should we expect them to be? Please consider strengthening this part of the manuscript. I could not glean much as presented.

There were nine patients with FLT3-ITD, and these samples are now indicated in Figure 4G, as well as more information about the stromal samples themselves. The primary stromal samples were collected from a variety of de novo and relapsed AML patients and therefore represents a heterogeneous mixture of AML patients at both diagnosis and relapse. There wasn’t a clear association of FGF2 and FGFR1 overexpression with FLT3 ITD, genetic markers, or previous treatment; but there were not enough samples to be conclusive. We did not directly compare expression levels to HS-5 and HS-27 by QPCR.

In our previous publications, the amount of FGF2 in CML and AML marrow core biopsies by immunohistochemistry was not significantly increased compared to normal marrow prior to treatment, and FGF2 only became significantly increased after treatment (Traer et al. Blood 2014, Traer et al. Cancer Research 2016). Therefore we would not necessarily expect more FGF2 expression in stromal samples from patients at diagnosis. We are currently culturing stromal samples from AML patients treated with a FLT3 inhibitor as part of a clinical trial to evaluate change in FGF2 expression over time, but do not have this data yet.

To us, the important feature of the primary stromal expression data is the coordinate upregulation of FGF2 and FGFR1 to drive autocrine activation in subsets of primary stromal samples. This would be predicted by our previous data, as well as published data on FGF2-FGFR1 autocrine activation in murine marrow stromal cells during stress hematopoiesis (Itkin et al., 2012, Zhao et al., 2012). Finding evidence for this autocrine loop in primary cells indicates it can be activated under the correct conditions. We are very interested to discover what signals from treated leukemia cells stimulates FGF2 and FGFR1 expression in marrow stroma, and this is currently under investigation.

The text was changed as follows:

“To evaluate FGF2 and FGFR1 expression in primary leukemia stroma, bone marrow aspirates from a series of leukemia patients were cultured ex vivo and FGF2 and FGFR1-4 expression quantified by RT-PCR (Figure 4F). FGFR1 and FGF2 transcripts were the most highly expressed in primary stroma, and there was a strong positive correlation between FGFR1 and FGF2 expression (Figure 4G, r2=0.5683 and p<0.0001 on nonparametric correlation). This indicates that FGF2 and FGFR1 expression are coordinately regulated in primary marrow stromal cells consistent with activation of an FGF2-FGFR1 autocrine loop. There were 9 stromal cultures from AML patients with FLT3 ITD (indicated with red dots), but most of them were newly diagnosed, and based upon our previous data we would not expect increased expression of FGF216. Similar to our observations in cell lines described above, we also detected FGFR1 and FGF2 in ECVs derived from primary marrow stromal cultures (Figure 4—figure supplement 2A). However, primary marrow stromal cells grow slowly and produce smaller amounts of ECVs, so we were unable to evaluate the effect of FGFR inhibitors on cell morphology, growth, and ECV production with primary marrow stromal cells.”

Minor Comments:

1) Figure 1: In panel A, what is the result if the same experiment is carried out with ponatinib (10-25 nM)? dasatinib (10-25 nM)? In panel B, with midostaurin (1 µM)? This comment is somewhat redundant with major point (3) above; here, I am requesting that the data are included, at least in the supplement.

Now included in Figure 1—figure supplement 1. Please see above for detail.

2) Why was 1 µM imatinib used? This concentration, at least at the 48 h timepoint, does not reduce proliferation (Figure 1A) to the extent that AC220 does in MOLM14 cells. Thus, it is difficult to determine the extent to which CM provides a protective effect. What happens if a higher concentration of imatinib is used, or if a more potent TKI is used at a concentration that matches the AC220 magnitude of reduced proliferation?

We used 1 µM imatinib primarily for comparison with our previous experiments. Now included in Figure 1—figure supplement 1 are a variety of ABL inhibitors and concentrations. Even with higher concentrations and more potent inhibitors K562 cells undergo apoptosis more slowly than MOLM14 cells.

3) Regarding the FGFR inhibitors:

– Why two? BGJ-398 is described as 'selective' but I did not find the rationale for needing it in the study (which began with PD173074).

– What is the direct evidence/readout to confirm that these inhibitors are working (and to measure the extent of inhibition)?

Previous reviewers requested experiments with more selective FGFR inhibitors and BGJ-398 is one of the newer and more selective inhibitors.

4) Perhaps it's not a reasonable comparison, but it is surprising that shRNA-based elimination of FGFR1 does not reduce FGF2 levels in HS-5 cells (Figure 6A), but FGFR inhibition does (Figure 5C). Also, the HS-5 CRISPR FGFR1.1 and 1.2 lines show reduced FGF2. The HS-5 CRISPR FGF2.1 and 2.2 lines show enhanced FGFR1 but reduced FGFR1 in ECVs. Can the authors comment on / explain these points?

We suspect most of the variability with FGF2 levels after shRNA or CRISPR/CAS of FGFR1 or FGFR inhibition are assay specific and related to when the cells were evaluated by immunoblot. Reviewer #1 also commented on this, and a more complete discussion is outlined above.

5) In Figure 6B, I would like to be able to evaluate whether reduced FGFR1 brings about a reduction in FGF2, but this data is not provided. In general, both FGF2 and FGFR1 should be shown in immunoblot stacks. Sometimes, it is one or the other.

Reviewer #1 also raised this issue and the absence of FGF2 staining in that figure was a technical issue (see above). Since we often had limited supplies of ECVs, we had to run the entire sample to obtain enough signal for detection, which did not allow for repeat immunoblots.

6) Figure 7:

– why is nilotinib used in 7A? All other data are with imatinib.

– the main point of the paper seems to be that interrupting the FGFR1-FGF2 network in stromal cells would decrease the protective effects emanating from the microenvironment. To test this, shouldn't arms with "PD PD173074" alone, and with "nilotinib + PD173074" be included?

– why did you choose this model rather than a FLT3-ITD model? The jumping back and forth between CML and AML is distracting. Could the authors consider a focus on MOLM14 and FLT3-ITD driven AML in the main manuscript, with the CML work allocated to the supplement?

We used the more potent inhibitor nilotinib to allow once daily gavage. This still allows for disease suppression and reduces trauma to the mice (and lab technicians). Imatinib is usually given twice daily by gavage.

The combination of an FGFR inhibitor and ABL inhibitor would be predicted to also reduce resistance, but the FGF2 -/- is a cleaner genetic model. We are currently trying to find pharma support to test relevant combinations with an FGFR inhibitor in clinical development.

The BCR-ABL model is a well-characterized model in the lab and does not require breeding. The FLT3-ITD and TET2 mouse model that we have access to would require extensive back-breeding with the FGF2 +/+ and -/- mice, adding significantly more time and expense.

We feel the focus of this manuscript on marrow stroma provides a unifying theme for our previous discoveries that FGF2 promotes resistance in both CML and AML. Our data also suggests that targeting the stroma directly with FGFR inhibitors is a novel therapeutic target in multiple leukemias.

7) Please address the following clerical issues:

– abbreviations are defined several times (examples: TKI; conditioned media)

– conventions are not consistently followed (ml vs. mL; h vs. hrs vs. hours; spacing between value and unit [e.g. 10nM vs. 10 nM]; no spacing vs. spacing preceding a citation)

– Chemical formulas lack proper subscripting (possibly an inherent submission portal limitation)

– Please check supplier names (Genscript or GenScript? Is Jackson laboratories the correct name of this supplier?)

– References 14 and 30 are identical

– A few article titles in References section are title case (#4, 15, 16, for example)

– Reference 23 lacks page number, etc. The final manuscript is 2016, not 2015

– Reference 25 lacks journal name, etc.

– Figure 2 legend: NB40 or NP-40?

– Final paragraph of subsection “FGF2 is Contained in Stromal Cell ECVs and Exosomes”: "membrate"

– First paragraph of subsection “FGF2-FGFR1 Signaling Promotes Stromal Growth and Paracrine Protection of Leukemia”: "withg"

– "proteinase K" vs. "Proteinase K"

– "PD173074" vs. "PD-173074"

– Second paragraph of the Discussion: a callout to Figure 3E is made. Should this be 3D (there is no panel 3E)?

Our apologies for the errors and inconsistencies. This has been corrected, although I left a definition of conditioned media in both the text and Figure 1 legend since I like to look at figures and legends first.

Additional data files and statistical comments:

I asked the authors to consider including additional data based on experiments with tyrosine kinase inhibitors beyond imatinib and AC220. If they choose to do this, inclusion of the data files is warranted. I do not see the need for any other additional data files.

The Prism data files were uploaded.

https://doi.org/10.7554/eLife.40033.029

Article and author information

Author details

  1. Nathalie Javidi-Sharifi

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Jacqueline Martinez
    Competing interests
    No competing interests declared
  2. Jacqueline Martinez

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft
    Contributed equally with
    Nathalie Javidi-Sharifi
    Competing interests
    No competing interests declared
  3. Isabel English

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
  4. Sunil K Joshi

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Contribution
    Data curation, Formal analysis, Investigation, Visualization
    Competing interests
    No competing interests declared
  5. Renata Scopim-Ribeiro

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Contribution
    Data curation, Investigation, Visualization
    Competing interests
    No competing interests declared
  6. Shelton K Viola

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Present address
    Department of Pediatrics, Division of Pediatric Hematology-Oncology, Naval Medical Center Portsmouth, Portsmouth, United States
    Contribution
    Data curation, Investigation
  7. David K Edwards V

    Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  8. Anupriya Agarwal

    1. Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    2. Division of Hematology and Medical Oncology, Oregon Health & Science University, Portland, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  9. Claudia Lopez

    1. Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    2. Center for Spatial Systems Biomedicine, Oregon Health & Science University, Portland, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  10. Danielle Jorgens

    1. Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    2. Center for Spatial Systems Biomedicine, Oregon Health & Science University, Portland, United States
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  11. Jeffrey W Tyner

    1. Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    2. Department of Cell, Developmental & Cancer Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    No competing interests declared
  12. Brian J Druker

    1. Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    2. Division of Hematology and Medical Oncology, Oregon Health & Science University, Portland, United States
    3. Howard Hughes Medical Institute, Chevy Chase, United States
    Contribution
    Resources, Supervision, Writing—review and editing
    Competing interests
    Is currently principal investigator or co-investigator on Novartis clinical trials. His institution, OHSU, has contracts with these companies to pay for patient costs, nurse and data manager salaries, and institutional overhead. He does not derive salary, nor does his lab receive funds from these contracts.
  13. Elie Traer

    1. Knight Cancer Institute, Oregon Health & Science University, Portland, United States
    2. Division of Hematology and Medical Oncology, Oregon Health & Science University, Portland, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    traere@ohsu.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8844-2345

Funding

American Cancer Society (MRSG-17-040-1 - LIB)

  • Elie Traer

Howard Hughes Medical Institute

  • Brian J Druker

National Institutes of Health (5F30CA186477-03)

  • Javidi-Sharifi Nathalie

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

Acknowledgements

We would like to thank Cristina Tognon and Kara Johnson for administrative assistance, and Angie Rofelty for assistance with primary bone marrow stromal cultures. Electron microscopy was performed at the Multiscale Microscopy Core (MMC) with technical support from the Oregon Health and Science University (OHSU)-FEI Living Lab and the OHSU Center for Spatial Systems Biomedicine (OCSSB).

Ethics

Human subjects: Bone marrow aspirates were obtained from AML patients after informed consent under the OHSU Institutional Research Board protocol IRB0004422.

Animal experimentation: Animal studies were carried out under approved OHSUInstitutional Animal Care and Use Committee, Protocol IP00000723. Number of animals for study and unnecessary suffering was minimized as much as possible.

Senior Editor

  1. Jeffrey Settleman, Calico Life Sciences, United States

Reviewing Editor

  1. Martin McMahon, University of Utah Medical School, United States

Reviewer

  1. Thomas O'Hare, University of Utah School of Medicine, United States

Publication history

  1. Received: July 12, 2018
  2. Accepted: January 16, 2019
  3. Version of Record published: February 5, 2019 (version 1)
  4. Version of Record updated: March 29, 2019 (version 2)

Copyright

© 2019, Javidi-Sharifi 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

  • 1,033
    Page views
  • 188
    Downloads
  • 0
    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. Cell Biology
    Helena Brezovjakova et al.
    Tools and Resources