High‐dose vitamin D metabolite delivery inhibits breast cancer metastasis

Abstract Besides its well‐known benefits on human health, calcitriol, the hormonally active form of vitamin D3, has been being evaluated in clinical trials as an anticancer agent. However, currently available results are contradictory and not fundamentally deciphered. To the best of our knowledge, hypercalcemia caused by high‐dose calcitriol administration and its low bioavailability limit its anticancer investigations and translations. Here, we show that the one‐step self‐assembly of calcitriol and amphiphilic cholesterol‐based conjugates leads to the formation of a stable minimalist micellar nanosystem. When administered to mice, this nanosystem demonstrates high calcitriol doses in breast tumor cells, significant tumor growth inhibition and antimetastasis capability, as well as good biocompatibility. We further reveal that the underlying molecular antimetastatic mechanisms involve downregulation of proteins facilitating metastasis and upregulation of paxillin, the key protein of focal adhesion, in primary tumors.


| INTRODUCTION
Multiple preclinical studies have demonstrated that vitamin D receptor (VDR), a nuclear receptor that modulates transcription of target genes, exerts profound influences on the initiation and progression of human cancers. [1][2][3][4][5] Vitamin D 3 (VD 3 ) from diets, supplements, or sun-dependent synthesis in lower layers of skin epidermis, is not the biologically active ligand for VDR. It undergoes enzymatical metabolisms first in the liver and then in the kidney to become calcitriol (Cal), which binds to and then activates VDR. 6 Breast cancer, as the most common cancer in women worldwide, in particular is highly related to VDR regulation. [7][8][9] Multiple evidence from preclinical studies have clearly shown that (1) low plasma levels of Cal increase risks of breast cancer and metastasis, [7][8][9] and (2) daily VD 3 supplement is helpful for breast cancer prevention. [10][11][12] The combinational treatments of Cal with cytotoxic chemotherapy, radiation, or other anticancer cytotoxic drugs showed additive or synergistic effects against breast cancer. [13][14][15][16][17] For example, the combined treatment of Cal or its analogs with gefitinib significantly enhanced the antiproliferative activity of gefitinib in EGFR and HER2 positive breast tumors. 14 On the other side, the question if Cal alone could be administered for breast cancer treatment, however, has not yet been systematically explored. Several studies showed promising but differing results. [18][19][20] The IC 50 of Cal was tested out to be around 10 nM on SUM-229PE and MCF-7 breast cancer cell lines during 6-day in vitro culture in the study led by Martínez-Reza et al. 20 Haddur et al. 21 found that IC 50  Excessive Cal administration can result in side effects including hypercalcemia, vascular calcification, and anaphylaxis, primarily because the drug shows nonselective distribution in vivo. [22][23][24][25][26] One of the major challenges with Cal for breast cancer investigations therefore is achieving a sufficiently high drug concentration within cancer cells. Delivering Cal via nanoscale drug delivery systems (DDSs) may overcome this, as DDSs have often proven successful at targeting drugs to tumors via active targeting mechanisms or passive targeting mechanisms like the enhanced permeability and retention effect. 27,28 DDSs carrying vitamin D or its derivatives, including succinic acidbased nanoparticles and liposomes, have recently been developed for cancer treatments, showing promising opportunities. 29 In this study, we develop a micellar nanosystem to deliver Cal specifically into breast cancer cells, we then systematically evaluate potentials of Cal on breast cancer treatment.

| RESULTS AND DISCUSSION
We synthesized a minimalist micellar nanosystem (mMNS) to deliver Cal specifically to breast cancer cells. mMNS self-assembled from cholesterol-PEG 5K (Chol-PEG 5K ) and cholesterol-DNA aptamer (Chol-DNA) conjugates. Differing from drug delivery systems prepared from laboratory-dominating special materials, both Chol-PEG 5K and Chol-DNA are widely and extensively used in research, and they are commercially available, which can thus facilitate the translation. Cal, as a fat-soluble compound, 30 can easily and stably be packed into the hydrophobic cholesterol core of mMNS (Figure 1a), forming the Cal containing mMNS-mMNS@Cal. The hydrophilic PEG 5K shell of mMNS@Cal can stabilize the system in vivo, prolonging its circulation in blood. 31,32 DNA aptamers targeting nucleolin on mMNS@Cal could direct the whole system into nucleolin-overexpressing breast cancer cells. 33,34 This system thus allows us to investigate pharmacological effects of high intracellular Cal dose on breast cancer.
We controlled the molar ratio of Chol-PEG 5K to Chol-DNA at 1 to prepare mMNS. With varying the concentration of each component, particle size and polydispersity index (PDI) measured by dynamic light scattering are used as the criteria to screen out the good preparation. Along with increasing the concentration of cholesterol conjugates or Cal, sizes of mMNS@Cal increased (Figure 1b), while corresponding PDI of each preparation also increased ( Figure 1c). This screening comparison leaded us to use 20 mg/ml cholesterol conjugates and 240.38 μM Cal as the final formula. The Cal encapsulation efficiency and loading efficiency with thus formula was measured to be 82.2% ± 2.6% and 32.8% ± 1.2%, respectively. mMNS@Cal prepared from this formula had its size at 42 ± 2.2 nm and PDI at 0.11 ± 0.03 nm, while the size and PDI of mMNS were 34 ± 5.1 and 0.09 ± 0.02 nm, respectively ( Figure 1d). Both mMNS and mMNS@Cal had similarly negative zeta potentials at around À1.5 mV, which should be mainly contributed by DNA aptamers around the systems ( Figure S1).
Transmission electron microscopy (TEM) imaging further validated our preparations ( Figure 1e). mMNS@Cal showed a good stability in the presence of serum at 37 C for 2 weeks, without big changes in size ( Figure S2). Less than 10% of loaded Cal was released from mMNS@Cal during this 2-week evaluation, further demonstrating its stability ( Figure S3). This slow releasing profile of Cal from mMNS@Cal in vitro might reduce the release of Cal in blood circulation, mitigating side effects caused by nonselective distribution of Cal.
After reaching tumor site and internalizing by tumor cells, the intracellular release profile of Cal would be accelerated by the lysosomes, 35 leading to an immediate drug release. To confirm the containing of DNA aptamer around mMNS@Cal, we incubated mMNS@Cal with a Cy5-modified complementary (to the sequence of DNA aptamer) DNA strand overnight under room temperature, which was then followed by washing unbound Cy5-modified DNA strand away via dialysis. The result showed that mMNS@Cal had strong Cy5 signal (data not shown), indicating the existence of DNA aptamer on it.
We firstly treated 4T1 breast cancer cells (with high nucleolin expression, which was confirmed by western blotting; Figure S4) with carboxyfluorescein (CFPE)-labeled mMNS@Cal (CFPE-mMNS@Cal) or nontargeting CFPE-mMNS@Cal (which was prepared as same as CFPE-mMNS@Cal but a random DNA sequence was used to replace the nucleolin-targeting DNA aptamer) (Figure 2a). We observed that CFPE-mMNS@Cal had a~30 times higher intracellular delivery than nontargeting CFPE-mMNS@Cal. Along with this, we compared the uptake efficacy of CFPE-mMNS@Cal on 4T1 cells and C166 endothelial cells (as nucleolin negative cell line, which was confirmed by western blotting; Figure S4). It again showed a~30 times higher CFPE fluorescent signal in 4T1 cells than it in C166 cells (Figure 2a,b). These observations indicated that DNA aptamers on mMNS@Cal can selectively mediate the nanosystem into cancer cells highly expressing nucleolin.
As the pharmacological result, IC 50 of mMNS@Cal (0.04 ± 0.01 μM) was much lower than IC 50 of Cal (>100 μM) or IC 50  Wound healing assay and transwell-based invasion assay are usually conducted to evaluate metastasis of cancer cells in vitro. 36 We thus carried out these two assays (Figure 2e,f) to study if high intracellular dose of Cal can impact the metastatic properties of breast cancer cells. At the concentration of 1 μM, Cal itself showed no effect on inhibiting the wound healing and cancer cell invasion. Apart from this, neither could the delivery system mMNS itself nor the mixture of mMNS and Cal significantly slow down the wound healing and transwell-based invasion. Nonetheless, mMNS@Cal showed great metastasis-inhibiting potentials on these two assay models. These in vitro experimental results together indicate that, at low concentration of 1 μM, Cal does have antimetastatic pharmacological effects, which however is restrained by its poor uptake by cells.
To assess if mMNS@Cal could specifically deliver high-dose Cal to tumor cells in vivo, we intravenously administrated it, at 5 mgÁkg À1 Cal, to mice orthotopically bearing 4T1 tumor. We then sacrificed mice at specific time points and quantified Cal in tumors and organs using high performance liquid chromatography (HPLC) measurement.
For Cal itself, at 4 h, it clearly showed that liver was the main biodistributing organ of Cal from both free Cal and mMNS@Cal administrations ( Figure 3a). There was no significant difference of tumorous Cal, which was around 2 μgÁg À1 , between free Cal and mMNS@Cal administrations. At 24th hour, tumors from mice treated with mMNS@Cal had Cal content at 5.61 ± 0.11 μgÁg À1 , whereas only 0.3 ± 0.11 μgÁg À1 Cal was tested out in tumors from mice treated with free Cal (Figure 3b). This around 18-fold increase of Cal in tumors indicated that mMNS@Cal can deliver high-dose Cal to tumors in vivo. Compared with free Cal administration, around 2.4 folds more Cal stayed in livers of mMNS@Cal treating mice at the 24th hour.
Cal, as the metabolite of VD 3 , is thought to be beneficial for health on many aspects. However, taking overdoses of Cal can lead to hypercalcemia by increasing the absorption of Ca 2+ in the kidneys, causing damages to bones, kidneys, heart, and brain. 37 We then evaluated the associated side effects of multiple times of Cal injection (5 mgÁkg À1 bw) on 4T1 tumor-bearing mice (Figure 3c). Compared to mice injected with Cal or the mixture of mMNS and Cal, the safety of mMNS@Cal treatment was highlighted by alleviated reductions of mice body weights (Figure 3d). More importantly, we measured blood Ca 2+ levels at different time points to assess risks of hypercalcemia. It showed that multiple times of Cal injection caused much higher blood Ca 2+ levels than the hypercalcemia threshold level (2.6 mmolÁL À1 ). 38 Nonetheless, under the same Cal dosages, mice injected with mMNS@Cal kept their blood Ca 2+ levels within safe ranges

| CONCLUSIONS
Thus, with all these results, it showed that our one-step selfassembled mMNS@Cal, with a high loading efficiency, can pack the hormonally active form of vitamin D into its cholesterol core. Directed by DNA aptamers modified around it, mMNS@Cal selectively delivered Cal into nucleolin-expressing breast cancer cells, both in vitro and in vivo, with high intracellular doses. We then proved that high intracellular doses of Cal reduced the growth of primary breast cancer and weakened metastasis. Our mechanism studies showed us that proteins facilitating tumor metastasis, including MMP-2 and MMP-9, were downregulated by mMNS@Cal. Besides, paxillin, as the key component of the focal adhesion complex, was upregulated by mMNS@Cal. These together implied that the molecular antimetastasis mechanism of mMNS@Cal is at least related to its ability to limit cancer cells' detachment from the primary tumor. More importantly, to fully unlock the anticancer and antimetastasis pharmacological functions of Cal, a reliable intracellular nanoscale drug delivery system like the system we used in this work will be needed.

| Blank micelles preparation
We used the thin-film hydration method as described in previous study to prepare micelles. 50

| Cal-loaded micelles
The workflow was same as the preparation of blank micelles except for the addition of Cal (amounts were indicated in Figure 1b

| Characterizations of the preparations
Size distributions were measured using a laser particle size analyzer (Malvern Nano ZS) with preparation suspension diluted in Milli-Q water at room temperature. Zeta potentials were investigated using a laser particle size analyzer (Malvern Nano ZS). Results were expressed as the average of three measurements. The size distribution is given by polydispersity index (PDI).

| Cell culture
Murine breast cancer 4T1 cells, luciferase-expressing 4T1 cells, and murine endothelial C166 cells were purchased from ATCC and cultured in RPMI 1640 Medium (Sigma-Aldrich). All cultures were supplemented with 10% fetal bovine serum (Gibco), 100 U/ml streptomycin, and 100 U/ml penicillin (Gibco) in a humidified atmosphere of 5% CO 2 at 37 C.

| Antimetastatic and antitumor assay
After tumor cell inoculation, we started treatments on the 10th day.
We randomly divided mice bearing luciferase-expressing 4T1 tumors into different experimental groups (10 mice per group). Preparations according to corresponding instructions. Ex vivo, three mice/group were sacrificed on the 20th day during the treatment and their tumors were collected. The tumors were diced into small pieces, and then lysed in cell lysis buffer containing protease inhibitors. After this, proteins in the tumor tissue lysis were analyzed by Elisa kits. 52 4.13 | Blood calcium level measurement During our treatment protocol, blood samples of mice (three mice per group) were collected from tail at specified days. Serum was then separated via centrifugation for calcium measurement. A fluorescencebased Calcium Assay Kit (ab112129; Abcam) was used, according to the product instruction, to read calcium levels in serum. 52

| Cal biodistribution
During our treatment protocol, after 24 h of the 6th drug injection, three mice per group were sacrificed and perfused (via heart) with cold PBS to collect organs. Organs were then homogenized within methanol. After centrifugation, supernatant was collected for Cal quantification by HPLC methods published before. 54

| Statistics
All the statistical analysis were carried out in R. When not otherwise stated, results were shown as mean values ± SD. Data were analyzed using two-tailed Student's t tests between two groups and one way analysis of variance followed by Turkey posttests among multiple groups. p value less than .05 were considered significant.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

DATA AVAILABILITY STATEMENT
The raw/processed data required to reproduce these findings cannot be shared at this time due to technical or time limitations.