Identification of an ALK-2 inhibitor as an agonist for intercellular exchange and tumor delivery of nanomaterial

Abstract

Inefficient extravasation and penetration in solid tissues hinder the clinical outcome of nanoparticles (NPs). Recent studies have shown that the extravasation and penetration of NPs in solid tumor was mostly achieved via an active transcellular route. For this transport process, numerous efforts have been devoted to elucidate the endocytosis and subcellular trafficking of NPs. However, how they exit from one cell and re-enter into neighboring ones (termed intercellular exchange) remains poorly understood. We previously developed cellular assays that exclusively quantify the intercellular exchange of NPs in vitro. Our study showed that a significant portion of NPs are transferred inside extracellular vesicles (EVs). Pharmacological inhibition of EV biogenesis significantly reduced the tumor accumulation and vascular penetration of both inorganic and organic NPs in vivo. Intrigued by this result, we performed here a manual chemical screen with our assay, which identified that LDN-214117 (an inhibitor for activin receptor-like kinase-2, ALK-2) is an agonist of NP intercellular exchange. We further showed that LDN-214117 regulates the intercellular exchange by increasing the EV biogenesis. Mechanistic investigation showed that LDN-214117 functions via BMP (bone morphogenetic protein)-MAPK (mitogen-activated protein kinase) signaling pathway to increase EV biogenesis. We further demonstrated that LDN-214117 treatment in vivo enhanced the tumor accumulation and vascular penetration of a variety of NPs in multiple tumor models, which improves their antitumor efficacy. Overall, we showcase here the identification of a novel chemical compound with our intercellular exchange assays to modulate EV biogenesis and EV-mediated transport, thus boosting up the delivery and therapeutic efficacy of nanomaterial.

Graphical Abstract

LDN-214117, an ALK-2 inhibitor, was identified as an agonist of nanoparticle transport from one cell to another (intercellular exchange) via extracellular vesicles (EVs). By increasing EV biogenesis via BMP-MAPK signaling, LDN-214117 enhances the overall accumulation and vascular penetration of nanoparticles in solid tumors of multilayered cells, and hence increases its antitumor efficacy.

Introduction:

Inefficient extravasation and penetration in solid tumor hinder the clinical outcome of nanomaterial^1–5, highlighting the importance to elucidate the underlying mechanism how they travel across the endothelium and through the tumor stroma composed of multilayered cells. Recent studies showed that an active transcellular transport route is responsible to carry most of nanoparticles (NPs) across the endothelium in solid tumors^6–8. In one study, Sindhwani et.al demonstrated that an active trans-endothelial pathway dominates the extravasation of NPs into tumors, while passive diffusion through intercellular gaps only accounts for a minor fraction⁷. After extravasation, macrophage was found to be responsible for transporting significant portion of NPs into deep tumor tissue ⁹. Another evidence arises from a study on the relationship between transcytosis of NP and its tumor penetration. Liu et.al found that enhanced transcytosis was responsible for the distribution and deep penetration of NPs in tumors⁸. The energy-dependent transcellular transportation process of NPs can repeatedly occur among cells in tumor tissue, resulting in the penetration of NPs into tumors and thus improving the therapeutic efficacy of NPs. Together, these results highlight the significance of the active transcellular transportation in NP delivery into solid tumor.

The process of transcellular transport can be generally divided into 4 steps: 1) the uptake of NPs into cells, 2) the intracellular trafficking of NPs, 3) the excretion of NPs by cells and 4) the re-entry of NPs into neighboring cells. While the first two steps have been extensively studied, the latter two are poorly understood^{10, 11}. We termed the exit from one cell (donor) and the re-entry into another (recipient) together as intercellular exchange, and developed cellular assays to monitor and quantify the intercellular exchange events in vitro^{12, 13}. In the first version of our assay, donor and recipient cells were simply mixed in ultra-low binding plates to form spheroids (termed “spheroid” model)¹². Our results with this assay revealed that a significant portion of NPs are transferred intercellularly in membrane-enclosed structures. We then adapted this assay into a collagen based three-dimensional (3D) assay¹³. With this assay, we confirmed that it is EVs, but not direct cell-cell contact, that are these membrane-enclosed structures for the intercellular exchange. We also demonstrated that EV-mediated intercellular exchange plays an important role in extravasation and penetration of NPs in vivo.

Cell culture systems beyond 2-D monolayer have already been integrated with high throughput screening (HTS) platforms^14–17. Therefore, we speculate that our assays can be used to screen for pharmacological agents regulating EV-mediated intercellular exchange of NPs, which can potentially improve the delivery and therapeutic efficacy of nanomedicine in solid tumors in vivo. To validate the feasibility, a pilot chemical screen was manually carried out, which identified LDN-214117 as the top hit.

Results

Identification of agonists for NP intercellular exchange through screening with our intercellular exchange assays

As a proof-of-principle study, we randomly selected 14 compounds from the Library of Pharmacologically Active Compounds (LOPAC^®1280) as shown in Table S1. All nanoparticles used in this studied were characterized as shown in table S2 by dynamic light scattering (DLS). The first round of screening was carried out in spheroid model as below. Cell penetrating peptide-conjugated silver nanoparticles (CPP-AgNPs) was first engulfed by donor cells (PC-3) via incubating cells with CPP-AgNPs. We chose CPP-AgNPs as model cargos for two major reasons: CPPs were found to facilitate internalization of NPs after binding with receptors on cell surface^{12, 18}, this would increase the number of NPs engulfed by donor cells. It is also demonstrated that CPPs can lead NPs from one cell to another^{6, 19}. These two features would increase the chance of intercellular exchange detected by our assay. As our previous work, the transactivating transcriptional activator (TAT)^20–22 and RPARPAR peptides²³ were chosen as the model CPPs to conjugate on AgNPs and used in this study. Another reason for us to use CPP-AgNPs is that the silver-based nanoparticles can be etched by a chemical and nontoxic solution (etchant)²⁴. This etchant cannot permeate the lipid membrane so that it can only remove extracellular AgNPs but not intracellular ones. By eliminating all those AgNPs outside cells, we can exclude all CPP-AgNPs transferring as free particles and focus on those transferred by EVs¹³.

The screening procedure was described in the Method section (Figure 1A). PC-3 (human prostate cancer cell line) and PC3-GFP cells were used to carry out the screening, as this cell line showed good susceptibility to CPP-AgNPs we used^{12, 13, 18}. TAT-conjugated AgNPs (T-AgNPs) and RPARPAR-conjugated AgNPs (R-AgNPs) were used as model CPP-AgNPs. Constant etching was used to focus the screening on EV-mediated NP transfer. Among all 14 compounds we tested, three compounds (CID 11210285 hydrochloride, LDN-214117 and ML277) significantly increased the intercellular exchange of T-AgNPs (Figure 1B), while two of them showed similar effect on that of R-AgNPs (Figure 1C). LDN-214117, an inhibitor for ALK-2, showed the highest stimulatory effects: the intercellular exchange efficacy of T-AgNPs and R-AgNPs was about 6.5- and 2.3-fold when compared to solvent control (DMSO), respectively.

Figure 1. Identification of LDN-214117 as an agonist for intercellular exchange of nanomaterial.

Schematic illustration of the screening process. (B and C) Compound screening in spheroid assay. Fourteen compounds from LOPAC were screened by spheroid assay as described in Methods. After 24 h incubation with the indicated compounds (x axis), relative intercellular exchange efficacy of T-AgNPs (A) and R-AgNPs (B) from PC-3 to PC3-GFP was quantified, respectively. (D-I) Validation of stimulatory effect of LDN-214117 on intercellular exchange of NPs in 3D assay. After 24 h incubation under the indicated conditions (x axis), intercellular exchange efficacy of T-AgNPs (D), R-AgNPs (E), T-Liposome (H) and T-PLGA (I) from PC-3 to PC3-GFP cells was quantified, respectively. The intercellular exchange efficacy of T-AgNPs (F), R-AgNPs (G) from HUVEC to PC3-GFP cells under the indicated conditions (x axis) was also quantified. And all results are normalized to that of control group (y axis). Error bars indicate S.E.M., n=3. *P<0.05, **P<0.01 and ***P<0.001 (Student’s t-test).

For hit validation, we tested LDN-214117 effect using our 3D assay. Since our main interest is to identify compounds that can stimulate EV-mediated transfer as well as tumor extravasation and penetration, the primary donor cell types included human umbilical vein endothelial cells (HUVECs) and PC-3. Besides T-AgNPs and R-AgNPs, TAT-conjugated Liposome (T-Liposome) and TAT-conjugated Poly Lactic-co-Glycolic Acid (PLGA) NPs (T-PLGA) were also tested as the examples of organic NPs. Liposome and PLGA NPs are two typical organic NPs that has been widely used as drug carriers^25–30. After confirming that LDN-214117 showed little cytotoxicity in 3D system (Figure S1), PC-3 and PC3-GFP were first investigated as donor and recipient cells. LDN-214117 significantly increased the intercellular exchange efficacy for both T-AgNPs (Figure 1D) and R-AgNPs (Figure 1E). Similar result was also observed for CPP-AgNP transfer between HUVEC and PC-3 cells (Figure 1F and G), as well as for the intercellular exchange of organic NPs (Figure 1H and I).

LDN-214117 regulates intercellular exchange via increasing EV biogenesis

Next, we set out to investigate the underlying mechanism of the LDN-214117 effect on intercellular exchange. First, we tested which stage of intercellular exchange LDN-214117 exerts its effect. To evaluate its effect on EV biogenesis, we quantified the protein amount of total EVs and NP-containing EVs produced by PC-3 cells upon LDN-214117 treatment. The particle concentration and size distribution of different types of EVs were also quantified by nanoparticle tracking analysis (NTA). As shown in Figure 2A–D, LDN-214117 significantly increased the total protein amount and particle concentration of both total EVs and NP-containing EVs when compared to control group, indicating that LDN-214117 increases EV biogenesis. Similar result was also observed in 4T1 breast tumor cells (Figure S2). No obvious alteration of size distribution of EVs was observed after LDN-214117 treatment (Figure 2E), suggesting that LDN-214117 does not regulate a specific subset of EVs. Meanwhile, we investigated the effect of LDN-214117 on NP uptake by donor cells, as well as EV re-entry into recipient cells. As shown in Figure S3, LDN-214117 treatment showed little effects on NP uptake by donor cell and the EV (originated from donor cells) uptake by recipient cells. To further validate this conclusion, we also measured the mRNA expression of genes that are known to belong to the machinery of EV biogenesis. Several genes, including ALIX, VPS4B, PTPN-23 and HGS, was significantly increased after LDN-214117 treatment (Figure S4). Overall, these results support that LDN214117 regulates intercellular exchange mainly via increasing EV biogenesis.

Figure 2. LDN-214117 regulates the NP intercellular exchange via increasing EV biogenesis.

(A-D) LDN-214117 stimulated the biogenesis of both total EVs and NP-containing EVs from donor cells. PC-3 cells were incubated w/wo LDN-214117 (x axis) for 24 h and let to engulf T-AgNPs. Total EVs and NP-containing EVs were isolated from PC-3 cells by density gradient ultracentrifugation. The protein amount from total EVs (A) and NP-containing EVs (B) were quantified (y axis) by BCA assay. The particle number of total EVs (C) and NP-containing EVs (D) were measured (y axis) by NTA. (E) LDN-214117 treatment didn’t change the size distribution of EV from donor cells. Size distribution profile of total EV originated from PC-3 w/wo LDN-214117 treatment by NTA. Error bars indicate S.E.M., n=3. *P<0.05, **P<0.01, ***P<0.001 (Student’s t-test).

To further confirm this conclusion, we collected conditioned medium from donor cells with or without LDN-214117 treatment. After removing freely released T-AgNPs in the conditioned medium by etching and ultrafiltration, we incubated recipient cells with conditioned medium for 24 h and the fluorescence intensity of T-AgNPs inside recipient cells was measured. As shown in Figure S5, the intracellular fluorescence intensity of recipient cells was significantly increased when incubated with conditioned medium from donor cells treated with LDN-214117. However, if EVs in the conditioned medium were lysed by Triton X-100 before adding back to recipient cells, no fluorescence can be detected in recipient cells. Taken together, these results confirmed that LDN-214117 induced elevated intercellular exchange indeed depended on the increasing production of EVs.

LDN-214117 increases EV biogenesis via BMP-MAPK pathway

Next, we set out to investigate the molecular mechanism that LDN-214117 increases EV biogenesis. Since LDN-214117 is an inhibitor of ALK-2^{31, 32}, we first validated that LDN-214117 effect depends on ALK-2 inhibition using an alternative method. We applied a small interfering RNA (siRNA) to inhibit the expression of ALK-2. After confirming the knockdown efficiency of ALK-2 siRNA (Figure S6), PC-3 cells treated by these siRNAs were used as donor cells to test the intercellular exchange efficacy of CPP-AgNPs. ALK-2 siRNA treatment showed little effect on the internalization of CPP-AgNPs by donor cells (Figure S7). We found that LDN-214117 can increase the intercellular exchange when donor cells are treated with negative control siRNA and ALK-2 siRNA, respectively (Fig. 3 A and B). However, LDN-214117 did not further increase the intercellular exchange when donor cells were pretreated with ALK-2 siRNA. Based on these results, we conclude that the increased EV biogenesis depends on the ALK-2 inactivation through either genetic or chemical approach, and this is the primary, if not only, mechanism for LDN-214117 effect. We also tested another ALK-2 inhibitor (LDN-193189) and observed similar results (Figure S8).

Figure 3. LDN-214117 stimulates biogenesis of EVs via BMP-MAPK signaling pathway.

(A and B) LDN-214117 increased intercellular exchange efficacy of CPP-AgNPs via ALK-2 inhibition. PC-3 cells were transfected with indicated siRNAs (x axis). The intercellular exchange of T-AgNPs (A) and R-AgNPs (B) from PC-3 to PC3-GFP cells w/wo LDN-214117 treatment were quantified, respectively. And the result is normalized to that of transfected with negative control siRNA without LDN-214117 treatment group (y axis). (C and D) Mechanism of LDN-214117 triggering EV biogenesis. PC-3 cells were incubated with indicated treatments (x axis) and let to engulf T-AgNPs and total EVs were isolated from PC-3 cells by density gradient ultracentrifugation. The particle concentration of EVs originated from PC-3 cells with BMP/LDN-214117 treatment (C) and with BMP/U0126 treatment (D) was measured by NTA (y axis), respectively. Error bars indicate S.E.M., n=3. *P<0.05, **P<0.01, ***P<0.001 and ns, no significance (Student’s t-test).

It is known that ALK-2 is a receptor of bone morphogenic proteins (BMPs)^{33, 34}. Thus, we speculate that LDN-214117 may regulated the biogenesis via BMP signaling pathway. We first tested the effect of BMP-2 protein on EV biogenesis. As shown in Figure 3C, BMP-2 alone decreased the amount of EVs secreted by PC-3 cells which is opposite to LDN-214117 effect. By pre-treating cells with LDN-214117, the inhibitory effect of BMP-2 on EV biogenesis diminished. Several works demonstrated that MAPK pathway can be activated by BMP-2 via SMAD-independent signaling pathway^35–37. K. Agarwal et al. found that the inhibition of MAPK signaling pathway can increase the number of exosomes released by cells³⁸. We hypothesized that LDN-214117 regulates the EV biogenesis by inhibiting the BMP-MAPK signaling pathway. To test this, we treated PC-3 cells with U1206, an inhibitor of MAPK, and found that U1206 showed similar stimulatory effect on EV biogenesis to LDN-214117 (Figure 3D). Together, these results suggest that LDN-214117 regulated EV biogenesis via inhibiting BMP-MAPK pathway.

LDN-214117 treatment enhanced tumor penetration and antitumor efficacy of nanomedicine

Based on in vitro results, we then tested whether LDN-214117 can be used to improve the tumor delivery and antitumor efficacy of nanomedicine in vivo. Here, we chose iRGD-functionalized liposome (iRGD-Liposome) and iRGD-AgNPs as the model NPs to first test the tumor homing and vascular penetration. Liposome is one of the most widely used NP-based drug carrier for cancer therapy and vaccination^{29, 30, 39} and iRGD is a peptide that can facilitate the vascular and tumor penetration of a variety of NPs^{40, 41}, including liposomes and AgNPs. LDN-214117 was first intratumorally injected into mice bearing 4T1 breast tumor every day for 5 days. The results of terminal deoxynucleotidyl transferase dUTP nick end labeling staining (TUNEL) confirmed that LDN-214117 induced little apoptosis in the tumor tissue (Figure S9). The homing study showed that LDN-214117 treatment significantly increased the accumulation of iRGD-liposomes in the tumor tissue (Figure 4 B). We also quantified the penetration distance of iRGD-liposomes from the nearest blood vessels, and found that LDN-214117 treatment can significantly increase the vascular penetration distance (Figure 4C). We observed a similar result when using an orthotopic pancreatic ductal adenocarcinoma tumor model (Figure 4D and E) and intraperitoneal injection of LDN-214117, indicating that systemic administration of LDN-214117 still has this effect. Besides iRGD-liposomes, we also tested the homing of iRGD-AgNPs and non-functionalized liposomes using 4T1 tumor model. As shown in Figure 4F and G, LDN-214117 can also increase the accumulation and penetration of iRGD-AgNP in tumor tissue. For non-functionalized liposome, while LDN-214117 showed little effect on the overall accumulation, it did significantly increase the vascular penetration (Figure 4H and I).

Figure 4. In vivo accumulation and penetration of NPs was improved by LDN-214117 treatment.

(A-E) LDN-214117 increased the accumulation and penetration of iRGD-liposome in 4T1 and orthotopic pancreatic ductal adenocarcinoma tumor model (KPC tumor model) in vivo. Tumor bearing mice received 5 daily injections of LDN-214117 (20 μL, 200 μM, intratumoral for 4T1 mice and 200 μL, 400 μM, intraperitoneal for KPC mice). 24 h after the last injection of LDN-214117, 200 μL of iRGD-Liposome was intravenously injected and circulated for 4 h. Tumor was excised and sectioned. Liposome and blood vessel were detected as described in Methods. (A) Representative images of iRGD-Liposome (FAM) and blood vessel (CD31) in 4T1 (left panel) and KPC (right panel) tumor model, respectively. Scale bar, 100 μm. (B and D) Quantitative analysis of relative iRGD-Liposome signal intensity in 4T1 (B) and KPC (D) tumor tissue, respectively. (C and E) Analysis of penetration distance of iRGD-Lliposome signal to the nearest blood vessel in 4T1 (C) and KPC (E) tumor by ImageJ. (F-I) Accumulation and penetration of iRGD-AgNP and Liposome in 4T1 tumor model in vivo with LDN-214117 treatment. 4T1 tumor bearing mice received 20 μL of LDN-214117 (200 μM) via intratumoral injection each day for 5 days. 24 h after the last injection, 200 μL of iRGD-AgNP / Liposome was intravenously injected and circulated for 4 h. Tumor was excised and sectioned for NPs and blood vessel detection as described in Methods. (F and H) Quantitative analysis of relative iRGD-AgNP (F) and Liposome (H) signal intensity in 4T1 tumor tissue by ImageJ. (G and I) Quantitative analysis of penetration distance of iRGD-AgNP (G) and Liposome (I) signal to the nearest blood vessel in 4T1 tumor by ImageJ (y axis). The in vivo experiment was carried out with 3 mice in each group. The signal intensity of nanoparticles were analyzed with 5 images from each tumor tissue. 45 Liposome signals or 90 AgNP signal were analyzed for the penetration distance. Error bars indicate S.E.M.. *P<0.05, **P<0.01 and ns, no significance (Student’s t-test).

Encouraged by these results, we set out to investigate whether enhanced delivery and vascular penetration mediated by LDN-214117 leads to a stronger antitumor efficacy of nanomedicine. Doxil (liposomal doxorubicin) is the first approved nanomedicine by FDA for cancer therapy and has been widely used in clinics³⁰. We functionalized liposomal doxorubicin with iRGD functionalization was previously shown to increase the tumor delivery and antitumor efficacy of liposomal doxorubicin (Sugahara et al, 2009). Here, we made iRGD-liposome-doxorubicin (iRGD-lipo-DOX) as the model nanomedicine. The treatment study was carried out using murine 4T1 breast subcutaneous tumor model on both flanks of mice as shown in Figure 5A. While LDN-214117 alone group showed little inhibitory effect against tumor growth, it significantly increased the anti-tumor efficacy of iRGD-lipo-DOX reflected by tumor volume (Figure 5B) and weight (Figure 5C). The immune-fluorescence staining of liposomes and blood vessels confirmed that LDN-214117 increased both accumulation and vascular penetration of iRGD-lipo-DOX in the tumor tissue (Figure 5D and E). The TUNEL staining showed that the apoptosis level in tumor tissue was significantly elevated in the combination group than that of iRGD-lipo-DOX alone group (Figure 5F and S9). Similar synergetic anti-tumor effect was also observed with a lower concentration of LDN-214117 (Figure S10). No significant difference of apoptosis levels was observed in healthy tissues across all groups (Figure S11). Taken together, LDN-214117 can improve the anti-tumor efficacy of nanomedicine by increasing its delivery and vascular penetration into solid tumors without increasing side effects.

Figure 5. LDN-214117 enhanced the antitumor effect of iRGD-lipo-DOX.

(A) Scheme of two flanked tumor mice and drug injection of the in vivo anti-tumor study. 5 days after tumor inoculation, 20 μL of LDN-214117 (1 mM) was intratumorally injected in the tumor on the right flank of mice everyday till the end of the study. 20 μL of DMSO solution was intratumorally injected in the tumor on the left flank of mice as control. 10 days after tumor inoculation, iRGD-lipo-DOX (1 mg/kg equivalent dosage of Doxorubicin) or PBS (as control) was injected through tail vein every 3 days for 3 times. Tumor volume were monitored every other day. At the end of the experiment, tumor tissue was collected, weighted and sectioned for staining as described in Methods. (B)Tumor volume during the study. (C) Tumor weight at the end of the study. (D) Quantitative analysis of liposome signal intensity in 4T1 tumor tissue by ImageJ, and normalized to that of control group (y axis). (E) Quantitative analysis of distance of liposome signal to the nearest blood vessel in 4T1 tumor by ImageJ. (F) Quantitative analysis of TUNEL signal in 4T1 tumor tissue by ImageJ, and normalized to that of control group (y axis). 5 mice were included in each group for the in vivo anti-tumor experiment. The analysis of fluorescent signal intensity was carried out with 5 images from each tumor tissue and 14 liposome signals from each image were applied to analyze the penetration distance. Error bars indicate S.E.M.. *P<0.05, **P<0.01, ***P<0.001 and ns, no significance (Student’s t-test).

Discussion

Penetration into tissues composed of multilayered cells is an important factor to determine the delivery and therapeutic efficacy of nanomedicine. We previously developed assays to quantify the intercellular exchange, a key and yet poorly understood part of NP transport between cells via the transcellular route. Here, we performed a pilot study to demonstrate whether our assays can be used for chemical screening of agonists for NP intercellular exchange. Our results identified an ALK-2 inhibitor, LDN-214117, as an agonist of EV biogenesis and thus NP intercellular exchange. Mechanistic investigations showed that LDN-214117 increased the biogenesis of EVs via inhibiting BMP-ALK2-MAPK signaling cascade. In vivo studies validated that LDN-214117 improves the accumulation and vascular penetration of a variety of NP types, leading to a stronger anti-tumor effect (Figure 5).

Many platforms have been developed to study the transcellular transport of NPs⁴². Transwell co-culture system is a simple and the most widely used model to study the transcytosis, cargo transport from one side of a cell to another. It has also been used to study NP transport across physiological barriers, such as blood brain barrier ^{43, 44} and gastrointestinal mucus⁴⁵. While this system covers the initial endocytosis, subcellular transport and export, it omits the cargo exchange between one cell and another, which is different from our assays. Zebrafish^46–48 and tumor-vascular-on chip models^49–52 have been developed and utilized to study the transport of NPs into tumor tissues. Although these models better mimic the real physiological conditions, customized instrument is often required, and the inherent complexity poses a challenge for the compatibility with mainstream screening platform. As a comparison, our intercellular exchange assays have several advantages. First, we focus on the intercellular exchange stage of trascellular transport, which is far less understudied than the initial endocytosis and subcellular transport. This also helps simplify the system. Second, our assay design and the usage of AgNPs allow us to exclusively quantify EV-mediated NP transfer. Last but not least, we envision that these assays can be easily adapted for high throughput screening, which is advantageous over in vivo or organ-on-chip systems. The main goal of current study is to verify this hypothesis, and to see whether the positive hit of such screening can be translated in vivo.

We chose compounds from LOPAC library as they are well characterized for their target genes and biological functions. The spheroid assay was used as it requires minimal modifications to move forward for the actual screening, and AgNPs and constant etching were applied to focus on EV-mediated transfer. Three out of 14 randomly selected compounds were identified as positive hits, out of which LDN-214117 shows the highest agonistic activity and can be validated with 3-D assay. While AgNPs were used for the screening, we showed that LDN-214117 effect can be extended to other NP types, especially organic ones (liposomes and PLGA), proving the generality of the screening output.

LDN-214117 is an inhibitor for ALK-2 with high degree of selectivity and low cytotoxicity^{32, 53}. ALK-2 is a type I receptor of BMPs, which are members of the transforming growth factor-beta (TGF-β)^{33, 34}. Recent studies showed that LDN-214117 demonstrates beneficial effects in preclinical models of ALK-2 mutant diffuse intrinsic pontine glioma^{32, 53, 54} and Fibrodysplasia Ossificans Progressiva³². Mihajlovic et al. found that the inhibition of BMP signaling pathway by LDN-214117 can reduces proliferating and migratory potential of non-small cell lung carcinoma cells³¹. But to our best knowledge, there has been no previous study indicating that BMP/ALK-2 signaling regulates EV biogenesis. Here, we first investigated which stage LDN-214117 promotes the intercellular exchange. LDN-214117 treatment promotes the cellular secretion of EVs in total and NP-carrying ones, while it has no effect on the cellular uptake of EVs or NPs. Moreover, LDN-214117 treatment increased the mRNA expression of exocytosis-related genes. Together, these results suggest that LDN-214117 increases the intercellular exchange of NPs by boosting up EV biogenesis in general.

To further understand the linkage between ALK-2 inhibition and EV biogenesis, we explored the downstream of BMP/ALK-2 signaling pathway. MAPKs play important roles in controlling cell survival and adaptation to extracellular stimuli⁵⁵ and can be activated through a non-Smad TGF- β signaling pathway^35–37. By inhibiting the constitutive MAPK activation, Agarwal et al. observed an increased number of exosomes released from cells³⁸. Intrigued by these results, we used an MAPK inhibitor, U0126, to treat the cells with or without BMP-2 stimulation. We found that while BMP-2 alone reduces the EV production, U0126 treatment in presence or absence of BMP-2 can increases the biogenesis of EVs to the similar level. The same result was seen with LDN-214117. These results support the notion that BMP-2, by activating ALK-2 and then MAPKs, reduces the EV biogenesis and thus EV-mediated intercellular exchange of NPs. Future mechanistic investigations may decipher the interplay among BMP/ALK-2/MAPK, EV biogenesis and intercellular communication, as well as the underlying physiological functions.

Last, our in vivo studies verified that our top hit in vitro is effective in vivo to improve the tumor delivery and antitumor efficacy of nanomedicine. LDN-214117 administration, both locally or systemically, increases the total amount and vascular penetration distance of organic and inorganic NPs in several solid tumors. While we mainly used iRGD-functionalized NPs, some improvement was seen with unfunctionalized NPs as well, supporting the generality of LDN-214117 effect. This finding is in line with our previous study showing that reduced intercellular exchange by an inhibitor of EV biogenesis decreases accumulation and penetration of NPs¹³. Finally, we used iRGD-coupled liposomal doxorubicin (iRGD-lipo-DOX) as a model nanomedicine for therapeutic evaluation. Our results showed that LDN-214117 is able to significantly increase the ability of iRGD-lipo-DOX to slow down the tumor growth and reduce the final tumor mass. It is worth noting that the dosage of iRGD-lipo-DOX used in our anti-tumor study is equivalent to 1 mg/kg of doxorubicin, which is much lower than other works ranging from 3 to 6 mg/kg^{56, 57}. While increasing the cytotoxicity in the tumor tissue induced by iRGD-lipo-DOX, which is likely the basis of the better therapeutic effect, LDN-214117 administration did not increase the cytotoxicity in healthy organs induced by iRGD-lipo-DOX. Additionally, LDN-214117 alone did not exhibit any anti-tumor effect at the dosage we were using in this study.

Overall, our study showcases the feasibility of using the intercellular exchange assays to identify novel chemical agents that regulate EV biogenesis and EV-mediated intercellular exchange of NPs. Such activity can be also translated to a higher tumor delivery and antitumor efficacy of nanomedicine in vivo. Future explorations, especially screening on a larger scale, may open up new avenues to advance our knowledge in EV biogenesis and intercellular communications, and improve the clinical outcomes of nanomedicine by regulating its transcellular transport.

Materials and methods:

Cell Lines and Cell Culture

Human prostate cancer cell line PC-3 and mouse breast cancer cell line 4T1 were purchased from American Type Culture Collection (ATCC CRL-1435, VA, USA). PC3-GFP and KPC cells was a gift of Dr. Erkki Ruoslahti’s lab. Primary Human Umbilical Vein Endothelial Cells (HUVECs) was gift from Dr. David K. Wood, University of Minnesota. Dulbecco’s modified Eagle’s medium (DMEM, cat. no. 16777–129, VWR international, LLC.) supplemented with 10% fetal bovine serum (FBS, cat. no. 35–011-CV, Corning), and 1% penicillin–streptomycin (10000 U/mL) (cat. no. SV30010, Thermo Fisher Scientific Inc.) was used for culturing PC-3, 4T1, PC3-GFP and KPC cells. Endothelial Cell Growth Medium-2 BulletKit (EGM-2, cat. no. CC-3162, Lonza Inc., ME, USA) was used for HUVECs. All cells were maintained in a 37 °C humidified incubator with 5% CO₂. For extracellular vesicle (EV) isolation, EV-free FBS was added in DMEM instead of regular FBS.

Preparation of nanoparticles

Detailed preparation and characterization method of all nanoparticles used in this study can be found in supplementary information.

Preparation of the etchant.

20× Etchant stock:

Solution A: 0.2 M of tripotassium hexacyanoferrate (III) (K₃Fe(CN)₆, Sigma, CAS# 13746–66-2) in DPBS (Hyclone, cat. no. SH30028.02).

Solution B: 0.2 M of sodium thiosulfate pentahydrate (Na₂S₂O₃ : 5H₂O, Sigma, CAS# 10102–17-7) in DPBS.

Both solutions were stored in dark at 4 °C. Solution A and B were freshly mixed at 1:1 (V/V) and diluted 20× by medium before use.

Compounds screen in spheroid model.

Compounds screening was carried out with spheroid model in ultra-low attachment 96-well plate (Corning, cat. no. 89089–826). Briefly, donor cells were incubated CPP-AgNP -contained medium at 37 °C for 4 h. After CPP-AgNPs were taken up by donor cells, all extracellular CPP-AgNPs as well as cell surface bound ones were dissolved by treating donor cells with etchant. These donor cells were then dissociated, counted and mixed with recipient cells which labeled with a fluorescence color (GFP). The cell mixture was added into in a low-binding plate with different compounds treatment under constant etching condition. After 24 h incubation, the cell spheroids were collected, dissociated into single cells. The intercellular exchange efficacy between donor and recipient cells was measured by flow cytometry using a BD FACS Calibur flow cytometer (BD Biosciences, San Jose, CA). The intercellular exchange efficacy was calculated as:

$$Intercellularexchangeefficacy=P_{AgNP-positive}\times M_{AgNP-positive}$$

where P_{AgNP–positive} stands for percentage of AgNP-positive recipient cells and M_{AgNP–positive} stands for mean fluorescence intensity in AgNP-positive recipient cells.

3D intercellular exchange assay.

When reached 70–80% confluency, donor cells were incubated with nanoparticle-contained medium at 37 °C for 4 h. After incubation with nanoparticles, cells were washed with etchant for 30 s followed by PBS wash for 3 times. Cells were trypsinized, counted and resuspended with medium for further use. Recipient cells were also counted and resuspended for further use.

The 3D intercellular exchange assay was optimized and carried out as our previous report¹³. Briefly, sterile 10× PBS, 1N NaOH, H₂O and 9×10⁴ recipient cells in 200 μL cell culture medium was mixed with Type I collagen (Collagen Type I, Rat tail high concentration, 8.91 mg/mL, ref. no. 354249, Corning), making the final concentration of collagen to 2 mg/mL and pH around 7.4.The mixture was added into 96-well plate and incubated in the cell incubator for 15 min for collagen polymerization. After the recipient layer polymerized, 30 μL of 2 mg/mL collagen gap mixture (prepared using the procedure above without cells) was added into each well and incubated at 37 °C for another 15 min. Then, 1.8 ×10⁵ donor cells in 200 μL medium (with/without compounds of interest) with etchant was added into each well and incubated for 24 h. After the intercellular exchange was completed, medium was removed and 1% collagenase (Sigma, cat. no. C9263–1G) in FBS free medium was added into each well to dissociate collagen. The mixture was centrifuged at 300 RCF at 4 °C for 10 min and cells were collected and fixed with 4% formalin. The intercellular exchange efficacy was measured by flow cytometry using a BD FACS Calibur flow cytometer. The intercellular exchange efficacy was calculated as:

$$Intercellularexchangeefficacy=P_{AgNP-positive}\times M_{AgNP-positive}$$

where P_{AgNP–positive} stands for percentage of AgNP-positive recipient cells and M_{AgNP–positive} stands for mean fluorescence intensity in AgNP-positive recipient cells.

siRNA transfection.

The ALK-2 siRNA and negative control siRNA were acquired from Thermo (Cat # 4390824 for ALK-2 siRNA and Cat # 4390843 for negative control siRNA). PC-3 cells were reverse transfected with 10-nM siRNA using Lipofectamine RNAiMAX transfection reagent (Life Technologies) for 48 h.

Real-Time RT-PCR

Total RNA was extracted from cells with TRI reagent (Sigma-Aldrich). cDNA was synthesized from total RNA using an iScript™ cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA). Quantitative amplification by PCR was performed using PowerUp SYBR Green qPCR Master Mix (Thermo Fisher Scientific) by a StepOne Real-Time PCR System (Applied Biosystem, Foster City, CA). The ΔΔCt method was used to calculate the results. Human tata-box binding protein was used as internal control and the primers for all tested genes were listed in Supplementary Table S3.

After 48 h transfection with ALK-2 siRNA, PC-3 cells were used as donor cells in 3D intercellular exchange assay as described in “Validation of LDN-214117 in 3D intercellular exchange assay” section.

EV Isolation, purification and quantification

EVs were isolated and purified by ultrafiltration followed with density gradient ultracentrifugation. Briefly, cells were incubated with nanoparticle-contained medium at 37 °C for 4 h. Cells were washed with etchant for 30 s followed by PBS wash for 3 times. Conditioned medium was collected after 48 h of incubation. EVs were isolated using ultrafiltration method as described previously. Briefly, cells and debris were firstly removed from the conditioned medium by centrifugation at 4 °C, 300× g for 10 min followed by 2000× g for 10 min. Then, the supernatant was concentrated with a Centricon Plus-70 Centrifugal Filter (Sigma, UFC710008) at 4 °C, 3500× g for 40 min. Reverse spin at 1000× g for 2 min was applied to collecte the concentrated medium. A series of gradient solution (40 % (w/v), 20 % (w/v), 10 % (w/v) and 5 % (w/v) solutions of iodixanol) were prepared by diluting OptiPrep™ (60 % (w/v) Sigma, cat. no. D1556–250ML) with 0.25 M sucrose, 10 mM Tris–HCl, pH 7.5 solution. The isolated EVs were suspended in 0.5 mL of 5% gradient and then layered on top of a gradient consisting of 10%, 20%, and 40% OptiPrep (3 mL for each gradient). Gradients were centrifuged at 100000g for 18 h at 4 °C using a SW 40 Ti rotor. Fractions of 1 mL were collected from the top of the gradient. Total EV fraction (fraction 7, 8, 10 & 11) and NP-containing EV fraction (fraction 10 & 11) were diluted in PBS (1:25) and centrifuged at 100000g for 90 min at 4 °C. The pellets were resuspended in cold PBS and were stored at 4 °C for further use.

The protein concentration of EV was measured using the Pierce™ BCA® Protein Assay Kits (Thermo Fisher Scientific Inc., cat.no. 23227). The particle number of EV was measured by NTA.

Conditioned medium collection and treatment.

PC-3 cells (donor cells) were pre-treated w/wo LDN-214117 (20 μM) for 24 h and let engulf T-AgNPs for 4h. After washing with etchant and PBS, donor cells were incubated with FBS-free medium for 24 h and the medium was collected. The medium was then treated w/wo 0.1% Triton X-100 followed by etching and ultrafiltration with Amicon Ultra 15 centrifuge filter (MWCO 10000Da, Sigma, cat.no. UFC901024).

Animals

All animal studies were carried out in compliance with the National Institutes of Health guidelines and an approved protocol (IACUC Protocol ID: #2012–38734A) from University of Minnesota Animal Care and Use Committee. The animals were housed in a specific pathogen-free facility with free access to food and water at the Research Animal Resources (RAR) facility of the University of Minnesota.

Balb/c mice were purchased from Charles River Laboratory (Wilmington, MA). 4T1 subcutaneous tumor model was established according to a reported protocol⁵⁸. Briefly, 1×10⁶ 4T1 cells were suspended in PBS and subcutaneously injected into both flanks of female Balb/c mice.

The orthotopic pancreatic ductal adenocarcinoma tumor model was established according to a reported protocol⁵⁹. Briefly, surgery procedure was performed to expose the entire pancreatic body together with spleen to the outside of the peritoneal cavity. 1×10⁶ KPC cell mixed withMatrigel (ref. no. 354234, Corning) were injected into the tail of the pancreas.

In vivo homing study

Once the average tumor volume reached 80 mm³ for 4T1 tumor, mice were intratumorally injected with 20 μL of 1% DMSO or LDN-214117 (400μM) in PBS daily for 5 days.

For orthotopic pancreatic ductal adenocarcinoma tumor, 14 days post-surgery, mice were intraperitoneally injected with 200 μL of 1% DMSO or LDN-214117 (400μM) every day for 5 times.

24 h after the last LDN-214117 injection, 200 μL of Liposome/iRGD-Liposome was intravenously injected. 4 h after the injection, animals were then anesthetized with Avertin and underwent heart perfusion before tumor tissue excision.

In vivo anti-tumor study

5 days after 4T1 tumor inoculation, 20 μL of 1% DMSO in PBS was intratumorally injected daily into tumor on the left flank of mice while 20 μL of LDN-214117 (1 mM) was intratumorally injected daily into tumor on the right flank of mice until the end of the study. 10 days after tumor inoculation, 200 μL of PBS or 200 μL of iRGD-lipo-DOX (dose equivalent to 1 mg/kg doxorubicin) was intravenously injected via tail vein every three day for 3 times. Tumor volume was measured with caliper every other day during the study. 48 h after the last i.v. injection, animals were then anesthetized with Avertin and underwent heart perfusion before tumor tissue excision. Excised tumor tissue was weighted and cryo-sectioned for staining.

Immunofluorescence (IF) staining

Immunofluorescence staining was performed on frozen tissue sections. Briefly, PBS containing 1% BSA and 0.1% Triton X100 was used to block the sections at room temperature for 1 h. The sections were washed three times with PBS followed by the incubating with primary antibodies (1:200 dilution in blocking buffer) at 4 °C overnight. The primary antibodies used in this study are as follows: rabbit anti-fluorescein/Oregon Green (Invitrogen) and CD31 (Rat Anti-Mouse, Thermo Fisher Scientific Inc., cat. no. MA1–40074). Secondary antibodies (1:200 dilution in blocking buffer) were then applied on slides and incubated at RT for 1 h. After washing with PBS, sections were stained with Hoechst for nuclei, mounted and covered with a coverslip. The sections were examined under fluorescence microscope EVOS M5000 (Thermo Fisher Scientific). The fluorescence intensity of liposome signal and penetration distance were analyzed by ImageJ.

TUNEL staining

The In Situ Cell Death Detection Kit (Roche Diagnostics GmbH, Mannheim, Germany) was applied for TUNEL staining. Frozen sections were incubated with TUNEL reaction mixture for 1 h at 37°C, and the nucleus were stained with Hoechst. The images were taken with EVOS M5000 fluorescence microscopy and the fluorescence intensity of TUNEL positive cells was analyzed by ImageJ software.

Statistical Analyses.

All statistical analyses were performed with the GraphPad Prism software. All data are presented as mean ± S.E.M. (standard error of the mean). No outliers were excluded in this study. P values lower than 0.05 was considered as statistical significance ( *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001). The methods of statistical analyses have been indicated in figure legends.