Chemokine releasing particle implants for trapping circulating prostate cancer cells

Prostate cancer (PCa) is the most prevalent cancer in U.S. men and many other countries. Although primary PCa can be controlled with surgery or radiation, treatment options of preventing metastatic PCa are still limited. To develop a new treatment of eradicating metastatic PCa, we have created an injectable cancer trap that can actively recruit cancer cells in bloodstream. The cancer trap is composed of hyaluronic acid microparticles that have good cell and tissue compatibility and can extend the release of chemokines to 4 days in vitro. We find that erythropoietin (EPO) and stromal derived factor-1α can attract PCa in vitro. Animal results show that EPO-releasing cancer trap attracted large number of circulating PCa and significantly reduced cancer spreading to other organs compared with controls. These results support that cancer trap may serve as a unique device to sequester circulating PCa cells and subsequently reduce distant metastasis.


Results
Chemotactic activities of different chemokines and growth factors. To explore of the idea of fabricating cancer traps for PCa cells, our first task was to identify the chemokines/growth factors that are potent in promoting PCa recruitment. Based on literature search, EPO 19 , SDF-1α 23 , CCL5 33 , VEGF-C 34 , CCL2 35 and CCL16 36,37 were selected as potential candidates. Using Transwell cell migration system, we first determined the chemotactic ability of EPO (100 U/ml), SDF-1α (100 ng/ml), CCL5 (100 ng/ml), VEGF-C (100 ng/ml), CCL2 (100 ng/ml), and CCL16 (100 ng/ml) using highly-metastatic KD cells and poorly-metastatic PC3 cells. The concentrations for each chemokines and growth factors with highest chemotactic activities were chosen based on the manufacturer's information. Our results have shown that, as expected, KD cells are more sensitive to all biomolecules than parental PC3 cells at any given concentrations. Moreover, SDF-1α and EPO are the most potent cytokines of inducing the migration of KD cells (Fig. 1A). Subsequent studies were carried out to demonstrate the effect of EPO (Fig. 1B) and SDF-1α (Fig. 1C) on the migratory ability of KD compared with PC3 cells in a dose-dependent manner. Slow release property, loading capacity and cell and tissue compatibility of HA particles. The chemokine loading capacity and releasing capability of HA particles were evaluated in vitro. The loading capacities of HA particles were 17.5 μg EPO/mg particles and 1.23 μg SDF-1α/mg particles when the initial loading concentration was 20.0 μg EPO/mg and 1.60 μg SDF-1α/mg particles, respectively. In an in vitro system we find that the amount of released EPO and SDF-1α reached to 55% (55 μg) and 63% (5.0 μg) loading capacity of HA particles within 4 hours. After 4 hours, EPO and SDF-1α-loaded HA particles released at a relative slower speed of 0.32% (0.32 μg)/hour and 0.08% (0.0062 μg)/hour, respectively (Fig. 3A).
We found that HA particles have no apparent toxicity up to 1 mg/mL in vitro (Fig. 3B). The tissue compatibility of HA particles was evaluated using an mice subcutaneous implantation model and PLGA particles were used as controls. After implantation for 2 days, we found that, in comparison to PLGA particles, HA microparticles prompted significantly less inflammatory cells accumulation (Fig. 3C). Furthermore, a low amount of CD11b + inflammatory cells accumulation was observed at the HA microparticles implantation and saline injection site, suggesting that HA particles have good tissue compatibility ( Supplementary Fig. 1).
In vivo assessment of cancer trap. To investigate the capability of cancer trap, EPO-loaded and SDF-1α-loaded HA particles were administered in the subcutaneous cavity. After particle implantation for 12 hours, mice were IV injected with NIR-labeled cancer cells. The distribution of cancer cells was then monitored via NIR imaging daily for up to 5 days.
Based on NIR fluorescent intensities, we find that the chemokine-loaded HA particle implant recruited significantly more highly-metastatic KD cells than poorly-metastatic PC3 cells (Fig. 4). The maximal recruitment of KD   intensities at the implant site decreased after Day 4 (Fig. 4A). On the other hand, the maximal KD cell recruitment of SDF-1α-releasing HA was found at Day 1-12,100,000 ± 4,000,000 AU/implant site (estimated 54,000 KD cells/ implant site). The numbers of cancer cells at SDF-1α-releasing HA implant site reduced substantially (~70%) at Day 2, increased slightly at Day 3 (estimated 31,000 KD cells/implant site), and then reduced with time (Fig. 4B). The above results suggest that EPO-releasing traps have the capability to capture and retain cancer cells at the trap implant site for up to 3 days.
To test the hypothesis, we compared the PCa cell recruitment efficiency between the PCa cell recruitment efficiency between SDF-1α-and EPO-loaded implants using the same animal model at Day 2 (36 hours after cancer inoculation). As expected, our results show that EPO implants recruited greater than 7X more KD cells than PC3 cells. In addition, SDF-1α implants attracted approximately 7X more KD cells than parental PC3 cells (Fig. 5A,B). Nevertheless, EPO implants appears to be slightly more efficient than SDF-1α implants by recruiting greater than 1.3 times more KD cells (Fig. 5A,B).

evaluation of the localization of pca cells inside or surrounding of cancer trap. The recruitment
of KD cells in cancer trap was examined histologically. GFP + KD cells were used in this study and GFP + KD cells were labeled with Vybrant DiD cell labeling dye with excitation and emission wavelengths of 649 nm and 670 nm, respectively. Tissue sections of EPO-loaded particle implants (EPO + HA) and particles alone (HA) were imaged for its NIR signals. We find significant NIR signals surrounding EPO-loaded particle implant sites (Fig. 6A). Most of the signals is localized at the interface of the particle implantation sites and surrounding host tissue. These results support the ability of EPO-loaded particles to enhance the recruitment of KD cells migration toward the implants. By overlapping NIR and fluorescent images, we determine that NIR signals coincide with GFP signals, suggesting the presence of live KD cells in and around the particle implants (Fig. 6A,B). We also find that there are significantly more NIR signals in the EPO-loaded particle implants (EPO + HA) than the particles alone (HA) (Fig. 6B). Finally, the numbers of KD cells in tissue sections were quantified. In agreement with earlier observation, we find the EPO-loaded particle implants (EPO + HA) attracted greatly than 3X more KD cells than chemokine-free particle implants (HA) and particle free tissue controls (Control) (Fig. 6C). Further studies were performed to determine whether the implantation of the EPO-loaded particles would increase EPO concentration in blood. Cy5-labeled EPO was used in the investigation. After implantation of a Cy5-EPO loaded particle implant for different periods of time (1-5 days), 10 μl heparinized blood samples were collected on daily basis. By measuring the fluorescent intensity of the blood samples, very low level of Cy5-EPO (<0.00002 unit or equivalent to 0.00001% of implanted EPO) was found in blood sample ( Supplementary Fig. 2). These findings support our hypothesis that the EPO released from the particle implants may not exert any physiological effect systemically.

impact of cancer trap on cancer cells metastasis.
To study the impact of cancer traps on cancer metastasis, EPO-loaded particle implants (EPO + HA) or particle free tissue control (Control) were implanted subcutaneously (2 implants per animal, 100 µl/site) on the back of mice. After inoculation of KD cells for 36 hours, animals were sacrificed, and all internal organs were imaged using Kodak in vivo imaging system (Fig. 7A,B). The fluorescent intensities of different internal organs were then quantified to reflect the progression of PCa-associated www.nature.com/scientificreports www.nature.com/scientificreports/ organ metastasis. Noticeably, we found that EPO particles implants significantly reduced KD cell accumulation in lung (51 ± 4%, p < 0.05) compared to control (lung: 90 ± 11%) within 36 hours (Fig. 7C). These results suggest that cancer trap device can mitigate the incidence of cancer cell metastasis via circulation. Histological study also uncovers that significant less KD cells were found in lung sections with EPO-loaded implants (EPO+HA; estimated 820±430 KD cells/mm 2 view field) than those with untreated control (Control; estimated 44 ± 50 KD cells/mm 2 view field) (Fig. 7D,E). This result supports that cancer traps may lure cancer cells away from circulation and indirectly reduce cancer spreading and/or metastasis. Our flow cytometry results show that there is 5.3% of KD cells are EPOR + cells while only 1.4% of PC3 cells express EPOR ( Supplementary Fig. 3). Such difference is not associated with EPO treatment, since the expression of EPOR among KD and PC3 cells do not change after 48 hours EPO incubation. These results support the hypothesis that the upregulation of EPOR on KD cells may not responsible for the preferential recruitment of the cells to EPO-releasing cancer trap.

Discussion
Metastasis remains a challenging clinical problem that accounts for the majority of cancer mortality. Tumor progression towards metastasis is governed by a complex multi-step process whereby tumor cells dissociate from their primary site of growth, invade surrounding tissues, intravasate into a blood vessel or lymphatic vessel, survive in circulation, adhere to and extravasate from the vessel and form a new tumor at secondary site 24,[38][39][40] . Our in vitro and in vivo studies have demonstrated that EPO and SDF-1α are the most capable of luring metastatic PCa. These results are supported by several recent observations. Specifically, EPO and SDF-1α have been shown to recruit PCa cells 19,41 . It should be noted that our study does not exclude the possibility that other types of cytokines and chemokines may also participate in the migration of PCa. For example, CCL2 facilitates PCa cell growth and bone metastasis 35 .
HA particles were employed to form the cancer trap based on these previous observations. Firstly, since hyaluronic acid dermal filler was approved by FDA in 2008, studies have shown that biodegradable HA has been proven to have good safety and cell/tissue compatibility 21,42 . Secondly, HA implants were found to induce minimal inflammatory responses due to their nonimmunogenic properties 43,44 . Thirdly, HA particles can be easily mixed with any chemokine in aqueous solution with denaturing its bioactivity minimally 45 . Fourth, degradation rate of HA particles is tunable via crosslinking density. Recent research showed that the maximum duration of HA-based dermal filler reaches up to 12 months and the increase of chemical crosslink density slows down degradation rate for short-term cancer treatment 44,46 . Finally, since HA particles don't respond to the changes of the environment, diffusion, degradation or osmotic pressure are believed to be the main factors to drive the release of drug and protein from HA particles 47 .
To examine the effect of the implants, we focus on the behavior of migrating tumor cells that travel from the circulatory system to distant organs or our cancer traps. Compared to the orthotopic model or intratibial Quantification of DiD + cells at the site of EPO-loaded particles (labeled as "EPO + HA", 98 ± 6 cells/mm 2 ), particles alone (labeled as "HA", 31 ± 7 cells/mm 2 ) and particle free tissue control (labeled as "Control", 1 ± 2 cells/mm 2 ). (n = 3) Data are present as mean ± SD. (Student's t-test, *p < 0.05 compared to HA.) Scale bar: 100 µm (white) and 300 µm (red). White arrows point to the interface between the skin tissue and the implant and the areas between two arrows are implanted sites.

Scientific RepoRtS |
(2020) 10:4433 | https://doi.org/10.1038/s41598-020-60696-x www.nature.com/scientificreports www.nature.com/scientificreports/ injection model, using the intravenous injection model as an experimental metastases model is more suitable to determine the efficacy of our cancer traps because of the reproductivity and the consistency of this model 48 . The intravenous injection model demonstrates the similarity of cancer cell distribution in mice and PCa patients, except in bone metastasis 6,49 . However, we create a chemokine-releasing implant made of soft natural material to prove our hypothesis that our cell-free cancer trap has the potential to compete with organ metastasis once the cancer cells enter the circulation system in vivo, after our previous study elucidates the mechanism of cancer migration from an in vitro artificial lymph node model 50 . Bone metastasis occurs in stiff bone structure with various types of cells (osteoclast, osteoblast) and chemokines (TGFβ, RANKL and VEGF) 51,52 . It is more complicated to cure bone metastasis than cancer extravasation from blood to soft organs. Since most of cancer cells migrate through circulation system or lymph nodes, in this study we emphasize the development of a universal and tunable cancer trap for different kinds of cancers.
Our findings on the preferential recruitment of circulating PCa cells to the implanted cancer traps are supported by the results shown in many previous publications 50, [53][54][55][56] . Briefly, human bone marrow stromal cells seeded porous hydrogel scaffolds are found to form a vascularized niche in mice and also to recruit human circulating tumor cells released from an orthotopic prostate tumor xenograft 53 . Electrospun fiber networks have been shown to trap cancer cells from human blood 54 . Metastatic lymph nodes are shown recently to promote the migration of KD cells but not PC3 50 . Porous scaffolds made of poly(lactide-co-glycolide) have been found to recruit breast cancer cells and subsequently reduce the tumor burden within solid organs 55 and animal survival 56 .
Our results show that the cancer trap not only recruits metastatic cancer cells but also decrease the accumulation of PCa cells in the lung and liver. These findings support that cancer trap implantation may be able to reduce PCa metastasis and, perhaps, prolong the survival of PCa patients. In fact, it is well established that patients with visceral metastases have higher mortality rate 49,57,58 . In addition, metastatic organs of PCa such as lung, liver, pleura and bone are more commonly found among patients 3,49 . By repeated administration of cancer traps, it is www.nature.com/scientificreports www.nature.com/scientificreports/ possible to trap a large number of circulating PCa in the subcutaneous space and, thus, indirectly delaying their spread throughout the body. In addition, using localized radiation and/or chemotherapy in conjunction with the cancer traps, a cancer therapy may be developed to eradicate metastatic PCa cells. Finally, it is well established that cancer takes years to become metastatic and long-term treatment might be needed to achieve the desired therapeutic outcome. For long-term treatment, it is possible to create a slow release hydrogel implant to extend drug release up to months 59,60 . Multiple treatment cycles with cancer traps on high-risk patients may be required yearly, along with frequent follow-up cares and monitoring. For late stage cancer patients, cancer trap can be used as a treatment option to eradicate migrating cancer cells during the localized treatments. Finally, the cancer trap treatment may not only avoid cancer recurrence but also work to actively remove circulating tumor cells during the treatment.

conclusion
We have developed a new cancer trap using injectable HA-microparticles that can release different chemokines/ growth factor to preferentially attract circulating and metastatic PCa cells. Among chemokines tested, EPO and SDF-1α are the most potent cytokines to recruit metastatic PCa cells in vitro. From an animal model injected with metastatic PCa cells intravenously, subcutaneously implanted cancer traps are found to able to attract significant amount of circulating PCa cells and further reduce the presence of circulating PCa cells in several visceral organs, including lung. These results support the potential of cancer traps used in patients with metastatic PCa to reduce or prevent the incidence of distant metastasis.

Material and Method
Materials. HA (sodium salt, 700KDa) was purchased from Lifecore Biomedical (Chaska, MN, USA). preparation of hyaluronic acid microparticles. HA microparticles were synthesized via a water-in-oil microemulsion process as described earlier with minor revision 61 . Briefly, HA aqueous solution (3 mL, 1.5 wt % in 0.2 M NaOH) was added dropwise into a 50 mL oil phase solution (isooctane + 0.2 M AOT + 0.04 M 1-HP) and then DVS (15 mM) under homogenization at 28,000 rpm for 5 minutes using OMNI GLH homogenizer (OMNI international, GA, USA). The reaction was allowed to continue under a vigorous stirring (2,200 rpm for 1 hour) at the room temperature. The reaction was then stopped with the addition of 3 ml of HCl (0.2 M). The HA microparticles were collected via precipitation in acetone. The crude HA microparticles were then washed consequently with deionized (DI) water, ethanol and acetone. The purified HA microparticles were completely re-dispersed in DI water and then lyophilized for further use. characterization of hyaluronic acid microparticles. To visualize the appearance of the microparticles, some of the HA microparticles were labeled with CF ™ 488 A dye (Biotium, Inc., Fremont, CA, USA) as described earlier 62 . The structure of HA particle is composed by a dense sphere shell and a loose interior to offer a space that contains aqueous solution [63][64][65] . The microparticle morphology was observed and the images were captured using a Leica fluorescence microscope (Leica Microsystems, Germany) equipped with a Nikon E500 Camera (8.4 V, 0.9 A, Nikon Corporation, Japan). The obtained images were used to determine the average size and distribution of particles using Image J 66 . The degree of swelling is also tunable by varying the crosslinking densities 67,68 . Chemokine loading capacity and release kinetic were determined in vitro as mentioned in previous publication 69 . Briefly, Cy5-labelling chemokine (EPO: 100 μg, SDF-1α: 8 μg) was loaded into freeze-dried HA microparticles (5 mg) by a "breathing-in" method 70 when most of water solution accumulates in the interior space of HA particles and at the same time the size of particles is swelling to bring more solution into HA particles 64,67 . Subsequently, 500 μL of PBS buffer was added on the top of the HA microparticles incubated at 37 °C. At a predetermined time, the supernatant was collected, and the release medium was replaced with an equal amount of the fresh one. The amount of the released chemokine, based on a calibration curve, was measured using a microplate reader (Infinite ® M200; Tecan Group Ltd, Switzerland). The cumulative release was calculated as the total amount of released chemokine at a specific time relative to initial loading amount. The in vitro cytotoxicity of the HA microparticles was determined by using MTT assay of 3T3 Swiss albino fibroblast cells (ATCC, Manassas, VA, USA) as described previously 71,72 . The in vivo toxicity of the HA microparticles was evaluated using a mouse subcutaneous implantation model as described earlier 73 . cell culture and migration assay. Early study has shown that DAB2IP gene knockdown in PC3 cell, a poorly-metastatic line, increases its metastatic potential (also called as DAB2IP-knockdown PC3 cells or KD cells) 24 . These stable cell lines expressing dual reporter genes (GFP and luciferase) were maintained in RPMI1640 medium (Invitrogen, Carlsbad, CA, USA) containing 5% PBS as previous described 24 . Migration assays were performed in Transwell dishes (Corning Costar, Cambridge, MA, USA) as described earlier 73  www.nature.com/scientificreports www.nature.com/scientificreports/ In vivo cancer cell recruitment. The animal study was designed to assess the capability of the particle implants to reduce cancer metastasis using an intravenous injection model. The intravenous injection model has been widely used as an experimental model to determine the metastasis efficacy of cancers, including PCa 26,31,32,48 . To assess the ability of chemokine-releasing HA microparticles for recruiting PC3 and KD cells, immunocompetent Balb/c mice (Taconic Biosciences, Rensselaer, NY, USA) was used in this study. The animal experiments were approved by the Animal Care and Use Committee (IACUC) at the University of Texas at Arlington in accordance with the Animal Welfare Act and Guide for the Care and Use of Laboratory Animals. Furthermore, animal procedures also comply with the Public Health Service "Policy on Humane Care and Use of Laboratory Animals". The animal procedure is summarized below. First, HA microparticles were mixed with different kinds of chemokines prior to in vivo experiments. Various groups of HA microparticles (9% w/v, 100 µl/implant site) were implanted on the back of animals via a 21-gauge needle. After particle implantation for 12 hours, PCa cells (5×10 6 cells/ animal) were injected intravenously (IV) into mice. In vivo cell migration was monitored using Kodak In-Vivo Imaging System FX Pro (Carestream Health Inc., New Haven, CT, USA) as described previously 19,76 . To determine in vivo cell distribution, at the end of the study, internal organs were isolated and their associated fluorescence intensities were measured based on the NIR images. Finally, the extent of cancer cell recruitment and "cancer trap" biocompatibility was evaluated histologically as previous described 19 . Statistics. All the data were evaluated using two-tailed student t-test and presented as mean ± standard deviation. The differences among each group were compared based on ANOVA and Tukey-Kramer test. Differences were designated as statistically significant when P ≤ 0.05 (Student's t-test).

Data availability
The datasets from this study are available from the corresponding author on reasonable request.