Lipoproteins LDL versus HDL as nanocarriers to target either cancer cells or macrophages

In this work, we have explored natural unmodified lowand high-density lipoproteins (LDL and HDL) as selective delivery vectors in colorectal cancer therapy. We show in vitro in cultured cells and in vivo (NanoSPECT/CT) in the CT-26 mice colorectal cancer model that LDLs are mainly taken up by cancer cells, while HDLs are preferentially taken up by macrophages. We loaded LDLs with cisplatin and HDLs with the heat shock protein-70 inhibitor AC1LINNC, turning them into a pair of “Trojan horses” delivering drugs selectively to their target cells as demonstrated in vitro in human colorectal cancer cells and macrophages, and in vivo. Coupling of the drugs to lipoproteins and stability was assessed by mass and raman spectrometry analysis. Cisplatin vectorized in LDLs led to better tumor growth suppression with strongly reduced adverse effects such as a renal or liver toxicity. AC1LINNC vectorized into HDLs induced a strong oxidative burst in macrophages and innate anti-cancer immune response. Cumulative anti-tumor effect was observed for both drug-loaded lipoproteins. Altogether, our data show that lipoproteins from patient’s blood can be used as natural nanocarriers allowing cell specific targeting, paving the way toward more efficient, safer and personalized use of chemo-and immunotherapeutic drugs in cancer.


Introduction
Among promising immunotherapeutic approaches in cancer are those targeting macrophages.
Macrophages are present in substantial amounts in most solid tumors and influence tumor growth or regression through inflammatory and metabolic switch (1,2). Macrophages, depending on their inflammatory status, can have a phenotype either tolerogenic (pro-tumoral) or cytotoxic (anti-tumoral).
Those that infiltrate the tumor are tolerogenic. Switching the polarization of those macrophages so they become cytotoxic is being investigated by many scientists (3). We recently demonstrated that the stressinducible heat shock protein-70 (HSP70) was abundantly secreted by tumor cells and favored pro-tumor phenotype in macrophages. Accordingly, depletion of HSP70 induced tumor regression via a massive intra-tumor recruitment of cytotoxic macrophages (4).
Concerning cancer chemotherapy, platinum-derived drugs such as cisplatin are major compounds in cancer treatment, particularly in colorectal cancer. However, besides high efficiency, cisplatin cytotoxicity causes cellular damage in healthy tissues, leading to adverse side effects such as renal and liver failure (5), pulmonary fibrosis or increased cardiovascular events (6), dramatically limiting the dose of cisplatin that can be used on patients, thereby limiting its efficacy.
More specific cell targeting, toxicity and solubilization issues have prompted the development of nanovectorization approaches. Hence, cisplatin encapsulation into liposomes is considered a promising strategy to reduce the amount of free cisplatin in the plasma and specifically delivering it to target tissues. While most of these attempts did not show significant toxicity reduction, a few studies successfully passed phase II clinical trials but still need phase III validation (7). Another approach to improve cisplatin efficacy in colon cancer treatment has been recently proposed through the creation of orally-administered squalenoylated nanoparticles loaded with cisplatin (8).
In contrast to these artificial lipid nanovectors, in this work we have explored natural lipoproteins (LDL and HDL) as nanocarriers for colorectal cancer drugs. We studied cisplatin and the potential immunotherapeutic HSP70-targetting drug (AC1LINNC) binding to purified native human lipoproteins, and the influence of such complexation on their selective transport to colorectal cancer cells or macrophages.

Pharmacodynamics parameters of LDL and HDL
We first determined the pharmacodynamics (PK) parameters of lipoproteins in Balb/c mice to choose the optimal time for tissue distribution assessment. Purified HDLs and LDLs were labelled with DOTA-Bodipy-NCS and injected i.v. into mice. Mean concentration time courses are presented in Figure 1A-B. PK parameters were relatively consistent between animals for HDL ( Figure 1A  The goodness-of-fit plots ( Figure 1C-D) shows a good correlation between observed and predicted concentrations, assuming that HDL and LDL followed a bicompartmental model. The major variability was linked to the volume of distribution in the central compartment (i.e., blood) (Supplementary table 3). The developed model enabled us to determine the optimal limited sampling strategy for our further experiments.

Tissue distribution and cancer cell specificity of LDL versus HDL in tumor bearing mice
For in vivo tissue distribution and cellular uptake of LDL and HDL, lipoproteins were labeled with radiolabeled 111 Indium ( 111 In) DOTA-Bodipy-NCS and injected i.v. into colorectal CT26 tumorbearing mice. This was reached 12 days after subcutaneous injection of CT26 cancer cells into the left flank of Balb/c mice. Labeled LDL and HDL in vivo distribution was followed by SPECT-CT and fluorescence microscopy. Lipoproteins uptake in the tumor and different organs (liver, spleen, heart, kidneys, bladder and blood) was visualized over 72h in the whole animal (Figure 2A-B). As expected, the highest levels of both LDLs and HDLs were found in the liver ( Figure 2G, Supplemental Data 1A).
However, 24h after systemic injections, lipoproteins could be visualized in the tumors, reaching a peak between 48-72 h ( Figure 2A&B). HDL and LDL increase in tumors inversely correlated with their decrease in the heart ( Figure 2C-F), indicative of lipoprotein decrease in the bloodstream. Of note, low amounts of both HDL and LDL were observed in the bladder throughout the experiment (Supplemental Data 1B), suggesting that both lipoproteins were not rapidly cleared in the urine. It is worth noting that LDL was more difficult to visualize within the tumor most probably because of its much faster clearance ( Figure 1B).
We next aimed to determine LDL and HDL subpopulations fates within the tumor. Twelve hours after injection of bodipy-bound HDL and LDL, the tumors were dissociated and submitted to FACS analysis ( Figure 3A&B). Interestingly, we observed that LDL-bodipy was preferentially up taken by CD45cells, which are predominantly tumor cells, while HDL-bodipy preferentially accumulated in macrophages (CD45 + /CD11b + /F4/80 + /Ly6G -, Figure 3B). This differential targeting of LDL versus HDL was confirmed in vitro using human cultured cells. Indeed, colorectal cancer HCT116 cells preferentially accumulated LDL-bodipy ( Figure 3C) while macrophages (from healthy volunteers' buffy coats) mainly up took HDL-bodipy ( Figure 3D) HDL as vectors for potential immunotherapeutic agents such as the HSP70-targetting

AC1LINNC.
With the rational that HDLs abundantly accumulated in macrophages, we decided to complex HDL to a molecule that we screened for its ability to target the heat shock protein-70 (HSP70), a chaperone known for its tumorigenic role involving macrophages differentiation/activity (4,(9)(10)(11). This molecule, designated as AC1LINNC (Supplemental Data 2A), binds to HSP70 with an IC50 of 0.2µM (Supplemental Data 2C) and inhibits HSP70 chaperone activity (Supplemental Data 2B&D).
AC1LINNC is highly hydrophobic and insoluble hampering its use in vivo, thereby increasing the interest for its vectorization. We incubated HDL or LDL (1mM cholesterol each) with AC1LINNC (to a final concentration of 100µM) (Supplemental Data 3A). Mass spectrometry analysis revealed a 30% uptake of AC1LINNC by both LDL and HDL allowing to achieve a final concentration of about 30µM ( Figure 4A). To evaluate the stability of AC1LINNC-bound lipoproteins and possible exchange of AC1LINNC between LDL and HDL particles, AC1LINNC-bound LDLs or AC1LINNC-bound HDLs (1mg/ml) were incubated with native HDLs or LDLs (1mg/ml), respectively (Supplemental Data 3B).
AC1LINNC was not detected in the newly added native lipoproteins, indicating that the interaction is stable with no exchange with other lipoproteins ( Figure 4B). We found that while the AC1LINNC complexed to LDL had no effect, the same amount of the AC1LINNC vectorized in HDL or solubilized in DMSO induced ROS production in macrophages ( Figure 4C).

Lipoproteins as nano-vectors for the chemotherapeutic drug cisplatin
We next used cisplatin, an efficient drug whose heavy undesirable effects hamper its use.
Lipoproteins were incubated with cisplatin (final concentration of 1mg/ml) for 4 hours at 37°C (Supplemental Data 3A). After dialysis, total cisplatin concentration in lipoproteins was measured by GF-AAS. We observed an integration of 30% and 50% of the initial dose of cisplatin, into LDL and HDL sub-fractions respectively ( Figure 5A). As previously described for the AC1LINNC (Supplemental Data 3B) no cisplatin exchange with other lipoproteins was detected ( Figure 5B). In sillico (Supplemental Data 4A&B and Supplemental table 4) and raman spectroscopy studies (See Supplemental Data 4C) indicated that cisplatin bound cysteine residues of the ApoB-100 protein.

Antitumor in vivo effect of cisplatin complexed to LDL
The impact of cisplatin-LDL versus -HDL complexes on in vitro tumor cell death was determined by MTT viability test. We observed that LDL-vectorization improved cisplatin-induced cancer cell mortality ( Figure 5C), while HDL-vectorization had no effect compared to the control (Supplemental Data 5A). These results were in agreement with our previous data showing that HDLs were not captured by HCT116 tumor cells while LDLs hardly entered macrophages. Consistently, we observed that cisplatin complexed to LDL had no effect on macrophage ROS production ( Figure 5D Figure 5E). LDL-Cis treatment was associated with a stronger tumor regression than cisplatin alone ( Figure 5F), while no significant effects of HDL-Cis or native lipoproteins were observed on tumor growth (Supplemental Data 5C). Consistently, we observed that LDL vectorization of cisplatin was associated with a higher tumor cell death (caspace-3, Figure 5G&I) and a higher proportion of macrophages infiltrating the tumor (F4/80, Figure 5G&J), compared both to control and non-vectorized cisplatin groups. Interestingly, cleaved-caspase-3 staining did not co-localize with F4/80, suggesting a selective pro-apoptotic effect on tumor cells. Additionally, this pro-apoptotic effect was associated with increased ROS production as assessed by DHE/DAPI staining ( Figure 5H&K). To summarize, LDL vectorization improves the efficiency of cisplatin, by increasing tumor cell death and, most probably indirectly, favoring cytotoxic macrophage infiltration.

Vectorization of cisplatin by LDL strongly diminishes cisplatin adverse side effects
To investigate the impact of lipoprotein vectorization on cisplatin nephrotoxicity, mice were treated with cisplatin using a standardized kidney injury protocol. Tumor-bearing mice were injected with a single dose of 30 mg/kg cisplatin and euthanized at day 3. Toxicity was assessed by evaluating weight loss and renal dysfunction ( Figure 6A). Mice treated with Cis-Pt experienced a 19% weight loss only 3 days following injection. In contrast, mice treated with LDL-bound cisplatin did not experience any weight loss ( Figure 6B), while, interestingly, no decrease in cisplatin anti-cancer drug efficiency was observed ( Figure 6C).
Following cisplatin treatment, mice displayed distortion of the overall renal morphology, dilation of renal tubules, and appearance of protein cast, most of which were significantly attenuated in the cisplatin lipoprotein-vectorized group ( Figure 6D&E). In addition to these observations, apoptosis assessed by caspase-3 cleavage was dramatically decreased in the kidney of mice treated with cisplatin complexed to LDL ( Figure 6F&G), while the induction of apoptosis in tumor cells was comparable in both groups ( Figure 6F&H). Furthermore, since the liver is involved in lipoprotein turnover, we aimed to assess if cisplatin complexed to LDL would induce hepatic toxicity. Interestingly, we found that LDLbound cisplatin did not induce hepatic toxicity, as shown by comparing apoptosis (cleaved caspase-3) in the liver of animals treated with cisplatin versus LDL-cisplatin (Supplemental Data 1B&C).
Altogether, cisplatin complexed to LDL, while preserving cisplatin anti-tumor effect, displays a reduced side effect toxicity.

Combinational effect of AC1LINNC-HDL and cisplatin-LDL complexes
Finally, we tested the impact of the association of cisplatin-LDL complexes together with AC1LINNC-HDL complexes. Tumor-bearing mice were treated with LDL-Cisplatin, HDL-AC1LINNC or the combination of both. While we did not observed a stronger decrease in tumor growth when using the combinational therapy ( Figure 7A&B), immunofluorescent staining revealed i) a strong induction of cancer cell apoptosis ( Figure 7C&E), comparable to that observed in the animals treated with LDL-Cisplatin alone ii) a strong burst in macrophage infiltration ( Figure 7C&F) comparable to HDL-AC1LINNC alone. This suggests that this combined strategy aiming to simultaneously target cancer cells with one drug (cisplatin-LDL) and tumor-infiltrating macrophages with the other (AC1LINNC-HDL) allows a complementary additive effect that could prove more efficient during prolonged treatment.

Discussion
In the field of cancer treatment, liposomal encapsulation of lipophilic drugs such as doxorubicin (Doxil, AmBisome) or vincristine (Marquibo) led to improved efficacy and safety. These new formulations were approved by the US Food and Drug Administration (12,13). However, the use of liposomal carriers for hydrophilic cisplatin has not achieved these goals so far. Among the few liposomal formulations of cisplatin that reached clinical trials, encapsulation of cisplatin in SPI-77® liposomes did not produce significant clinical efficacy in phase II studies despite its safer toxicity profile (14), due to inefficient release of the drug from the carrier (15). Although Lipoplatin® (active encapsulation of cisplatin into PEGylated liposomes) showed enhanced antitumor efficacy compared to free cisplatin in some patients along with reduced renal toxicity, phase II and III studies demonstrated inconsistent effects on survival rates (16). Finally, LiPlaCis®, a phospholipase A2-sensitive liposomal cisplatin carrier have shown a poor safety profile, leading to the discontinuation of phase I studies (17).
In the present study, we considered native human LDL and HDL as carriers for several reasons: 1) lipoproteins are endogenous, physiological and stable molecular complexes that can carry a wide variety of molecules in the bloodstream; 2) they bear specific apolipoproteins that allow targeting to specific receptors; 3) certain chemotherapeutic drugs such as cisplatin display affinity for plasma proteins and therefore are likely to bind to apolipoproteins (18). This latter hypothesis was supported in the present work by in silico studies showing that hydrophobic cisplatin interacts with the protein moiety, and not the lipid moiety of lipoproteins. Further, we demonstrate in vitro that cisplatin association to LDL and HDL is highly efficient and stable in aqueous solution, with no subsequent release of bound cisplatin into the buffer or towards other non-loaded lipoproteins. These results, together with our in vivo kinetics experiments suggest that cisplatin binding to lipoproteins can significantly prolong its half-life in the blood (only 0.24 hours for free cisplatin (19)). This may allow the molecule to reach its target tissue instead of being quickly cleared in an unspecific manner.
Accordingly, we were able to show that substantial amounts of labeled lipoproteins could be found within CT-26 tumors after their systemic administration in mice.

Interestingly, LDL and HDL targeted different cells within the tumor. While labeled HDL
preferentially accumulated in tumor macrophages, LDL were mainly found within CT-26 cancer cells.
A possible explanation is that LDL receptor (LDL-R) expression is abnormally elevated in cancer cells (20,21). In contrast, HDL, which uptake cholesterol excess from peripheral cells back to the liver, have been shown to interact with immune cells via specific membrane transporters and scavenger receptors, particularly present in macrophages. The specific tropism of the LDL for tumor cells may explain the practically lack of cisplatin's side effects such as renal and hepatic toxicity, when conjugated to LDL.
Since HDL particles are abundantly taken up by macrophages, we believe that loading of HDL with immunotherapeutic compounds may be more promising than with chemotherapeutic drugs. Thus, HDL might have a strong therapeutic interest to vectorise molecules that influence macrophages polarization toward an anti-tumor phenotype, and proof of the concept is shown in this work with an HSP70 inhibitor affecting macrophages differentiation/activation (9, 10). Overall, we conclude that the usage of HDL might allow in the future a more targeted/safer administration into the patient of immunotherapeutic agents alone or in combination with LDL-vectorized chemotherapeutic agents, such as cisplatin.

Cell culture and mice
HCT116 cells (CCL-247 TM ) were from the American Type Culture Collection (ATCC). Human macrophages were isolated from healthy donor Buffy Coats (EFS Besançon, France) and differentiated as described (22)

Cisplatin and AC1LINNC incorporation in lipoproteins:
Cisplatin (Sigma Aldrich, 479306-1G) was dissolved in 0.9%NaCl to a 10mg/ml concentration. with native HDL or LDL (1µM). Cisplatin incorporation in LDL and HDL was then assessed by GF-AAS (graphite furnace coupled to atomic absorption spectrophotometry) and AC1LINNC incorporation by mass spectrometry.

Bio-distribution and pharmacokinetics imaging
Tumor bearing Balb/c mice (~300mm 3 ) received 5µg 111 In-DOTAGA-HDL or 111 In-DOTAGA-LDL (8−10MBq) i.v.. SPECT/CT dual imaging was performed 1h, 24h and 72h after injections using a NanoSPECT/CT small animal imaging tomographic γ-camera (Bioscan Inc., Washington, DC). Mice were anaesthetized with isoflurane (1.5−3% in air) and positioned in a cradle. CT (55 kVp, 34 mAs) and helical SPECT acquisitions were performed in immediate sequence. Both indium-111 photopeaks (171 and 245keV) were used with 10% wide energy windows. Radioactivity was measured (from blood, tumor, and organs) with a scintillation γ-counter. Data were then converted to percentage of injected dose per gram of tissue (%ID/g). SPECT/CT fusion image was obtained using the InVivoScope software (Bioscan Inc.). Radioactivity contents from image analysis were expressed in Bq/g, converted to percentage of injected dose, and compared to those determined by ex vivo counting.

Statistical analysis
Differences among two groups were determined using two-tailed unpaired Fisher's t-test, and ANOVA followed by Bonferroni's multiple comparison test for more than two groups, using SigmaStat version 3 (GraphPad Software). Differences were considered significant when p < 0.05.

The work was supported by a French Government grant managed by the French National
Research Agency under the program "Investissements d'Avenir" with reference ANR-11-LABX-0021 (LabEX LipSTIC). We also thank the Conseil Regional de Bourgogne-Franche Comté, the European Union program FEDER, and the "Ligue National contre le Cancer" for their financial support.

Acknowledgments
We thank the Platforms Cellimap for their technical support. S.Y, S.