Endogenous N-oleoylethanolamine Crystals loaded Liposomes with Enhanced Hydrophobic Drug Loading Capacity for Efficient Stroke Therapy


 Background. Although the preparation of liposomes achieves great success and many have received regulatory approval, their retention of highly hydrophobic drugs is still problematic. Results. Herein we report a novel strategy for efficiently loading hydrophobic drug to liposomes for stroke therapy. N-oleoylethanolamine (OEA), an endogenous highly hydrophobic molecule with outstanding neuroprotective effect, was successfully loaded to OEA-SPC&DSPE-PEG liposomes (OSDP LNPs) with a drug loading of 15.9 ± 1.2 wt%, four times higher than those prepared via traditional methods. Efficient retention in OSDP LNPs greatly improved the pharmaceutical property and therefore enhanced the neuroprotective effect of OEA. Through the data of positron emission tomography (PET) and TTC-stained brain slices, it could be clearly visualized that the acute ischemic brain tissues were preserved as penumbral tissues and bounced back with reperfusion. The in vivo experiments stated that OSDP LNPs could significantly improve the survival rate, the behavioral score, the spatial learning and memory ability of the MCAO (middle cerebral artery occlusion) rats. Meanwhile, the cerebral infarct volume, the edema degree, the apoptosis of the neurons and the inflammation within the brain were also greatly decreased. Conclusions. These results suggest that the OSDP LNPs have a great chance to develop hydrophobic OEA into a potential anti-stroke formulation.

e ciency of chemotherapeutic agents by delivering high concentrations of the drug to tumor sites, while reducing their accumulation in normal tissues to minimize the side-effect [19]. Hence, the drug loading property of the liposome is one of the key factors which in uence their e ciency. However, the retention properties of drugs in liposomes are largely drug dependent [20]. Loading the drug in response to transmembrane pH gradients would further increase the drug loading and the stability [21]. Via the technology, weak-base drugs, originally di cult to retain in liposome, could receive a relatively high retention property [22,23]. However, retention of highly hydrophobic drugs in liposomes is still problematic [24][25][26]. For instance, the loading e ciency of paclitaxel by different methods varies from 1.5-4% by weight [27,28]. This extremely low drug loading is due to the highly hydrophobic property of paclitaxel [29]. Current strategies to develop liposome-based formulations of highly hydrophobic drugs have focused on incorporation of these drugs in the lipid bilayer of liposomes [20]. However, these formulations could not possess a high drug loading, and drug rapidly exchanges from the liposome bilayer, leading to their instability in the blood circulation [24].
To enhance the drug loading and the stability of liposomes, our group developed hydrogen bond based liposomes of various drugs [30][31][32][33]. Through forming hydrogen bond between the drug and soybean lecithin, the drug could be e ciently and rmly loaded in liposomes, possessing a high drug loading and enhanced controlled release property. Based on this method, we report a novel strategy here which could cleverly combine the self-assembly of the amphipathic molecules with the crystallization of the drug.
With the crystallization of hydrophobic OEA inside the OSDP LNPs, the drug loading increased by four times compared to those of traditional methods. In the in vivo experiments, the OSDP LNPs possessed an outstanding neuroprotective effect and achieved enhanced therapeutic e ciency in the treatment of the cerebral injury caused by MCAO.

Preparation and characterization of OSDP LNPs
A key feature of the research is the precursor mediated, controlled drug crystallization inside the liposomes. In order to obtain nanosized crystals in the liposomes, the traditional nanoprecipitation technique were optimized. The hydrogen bond were formed between OEA and lecithin (SPC) via our previous method [18,20], which was con rmed by H 1 NMR and XRD. In the H 1 NMR spectra (Fig. 1A), the peak at 7.76 ppm (ascribed to the -NH-of OEA), the peak at 4.63 ppm (ascribed to the -OH of OEA), and the peak at 3.47 ppm (ascribed to phosphatidylcholine of SPC) appeared to obviously weakened in the OEA-SPC complex, illustrating the formation of hydrogen bonds. The disappear of the sharp peaks of bulk OEA in the XRD pattern of OEA-SPC also con rm that the hydrogen bonds might have formed between OEA and SPC molecules (Fig. 1B).
Afterwards, extra OEA and OEA-SPC were dissolved in dichloromethane, which was then dispersed in DI water containing DSPE-PEG under ultrasonic conditions. After stirring for one hour, the system turned into a stable, white O/W suspension with the dispersion of the organic droplets in the continuous phase. In this stage, there was an arrangement of DSPE-PEG molecules with the hydrophobic groups of DSPE around the organic droplets and the PEG groups in the continuous phase. The droplets could be seen as precursors, which might proceed via non-classical crystallization routes. Lastly, with dichloromethane rapidly removed by rotary evaporation, the multistage crystallization process proceed. Due to the evaporation of the organic phase, water molecules entered the droplets, resulting in the precipitation and arrangement of SPC-OEA complex. Hence, phospholipid bilayers were prepared according with the arranged DSPE-PEG molecules. The bond of OEA to SPC molecules would serve many nucleation sites, leading to a number of OEA crystal nucleus. The rapid evaporation of the organic phase would induce a high supersaturation, under which the OEA crystal nucleus grew into OEA nanocrystals and rapidly reached the thermodynamically stable crystalline form. Since the phospholipid bilayers were exible, their shapes might be changed by OEA nanocrystals. Followed by extrusion through polycarbonate membranes (0.22 µm pore diameter), irregularly non-spherical shaped liposomes, with drug crystal inside and phospholipid bilayers outside, were successfully prepared. The representative structure illustration of the OSDP LNPs is depicted in Fig. 1D.
As illustrated in Fig. 1C and Figure S1, the OSDP LNPs were around 100 nm with an irregularly nonspherical shape. It could be noticed that the black elds inside of the liposomes appear darker than the elds forming the contour of the liposomes, which appear to be a paler shade of grey ( Fig. 1C-b and c). The inside dark eld might owe to the existence of OEA crystals, whose molecules arranged well-ordered and closely. Since the electrons could not easily pass through the closely packed crystals, the inside eld of the liposomes showed a deep color. On the contrary, the phospholipid bilayers were loosely arranged, leading to a higher electron transmittance and hence a light shade. The two different colored sections also illustrated the delicate hierarchical architectures of the OSDP LNPs.
X-ray diffraction was also employed to detect the form of OSDP LNPs. As shown in Fig. 1B, the sharp peaks of bulk OEA would disappear with the formation of hydrogen bonds in the OEA-SPC complex.
Nevertheless, a majority of sharp peaks of OSDP LNPs belongs to OEA, suggesting its high crystallinity. Although, the OEA within OSDP LNPs shows the same polymorphic form as their bulk counterparts, the width of the peaks have reduced a lot, which might attribute to the nanoscaled crystal size of OEA within OSDP LNPs. Meanwhile, the coverage of lipid bilayer on the OEA nanocrystals led to the weakness of OEA.

Stability test and in vitro drug release study
The biggest advantage of the OSDP LNPs lies in the sustained drug release property with a high drug loading. The drug loading of OSDP LNPs could reach up to 15.9 ± 1.2 wt%, while that of the OEA-SPC LNPs collected from the traditional method ranged from about 0.5 wt% to 3.0 wt%. With the formation of hydrogen bonds, the drug loading increased to about 8.0 wt%. When the nanoscaled OEA crystal appeared within the liposomes, the drug loading was redoubled to 15.9 ± 1.2 wt%. Although the drug loading of many liposomes might come up to more than 50 wt%, the retention of highly hydrophobic drugs is still a critical bottleneck. 20 The drug loading of many liposomes was less than 10 wt%, or even less. In this research, the formation of hydrogen bonds, plus the crystallization within the liposomes, greatly enhanced the retention of OEA with highly hydrophobic property. A well-established study showed that drug release could be controlled by the drug-loading content of the particles. High drug loading means severe burst release, while poor drug loading brings well controlled drug release [21] . As to OSDP LNPs ( Figure S2), higher drug loading also led to a little burst release within the rst 8 h, which might come from the exposed OEA crystals and the OEA on the surface of the liposomes. Compared to the OEA crystals without the coverage of lipid bilayer, this slight burst release could be neglected. Just like the OEA-SPC NPs with a drug loading of 8.2 wt%, the OSDP LNPs exhibited a remarkably prolonged and controlled drug release in the next 40 h, which could almost be seen as zero-order release. That is to say, the enhanced drug loading did not reduce the controlled release property of the liposomes. This was owing to the unique construction of the liposomes. Unlike other drug crystals, the OEA crystals was located in the center of the OSDP LNPs, covered with lipid bilayers and bonding with the SPC molecules, which would largely limit the diffusion of the drug. Hence, the OSDP LNPs possessed a steady sustained release pattern throughout the release period. Moreover, the lipid bilayers, the PEG chains, the monodisperse, the high Zeta potential (-16.1 mV, Figure S3) and the small particle size contributed to the stability of the OSDP LNPs. With minor size changes, the OSDP LNPs could preserve stability in an aqueous environment for 48 hours ( Figure S4). On the contrary, the OEA-SPC NPs (d = 213 nm) without coating of PEG chains would aggregate within 6 hours. And the lyophilized OSDP LNPs redissolved easier than OEA-SPC NPs. The redissolved dispersion of lyophilized OSDP LNPs possessed a size of about 180 nm and was still applicable for intravenous injection within 1 month, while the OEA-SPC NPs couldn't ( Figure S5). The reason was the strong hydrophilia of PEG, which could stabilize the liposomes and make them more soluble in water.

Assessment of the neuroprotective effect of the OSDP LNPs in vivo.
To assess the in vivo neuroprotective effects of the OSDP LNPs on ischemic cerebral injury, systematic experiments were performed on rats. According to our previous study [30,35], OEA could obviously reduce the cerebral infarct volume of the brains from the MCAO rats. Hence, it could be speculated that a lot of ischemic tissues were preserved to be ischemic penumbra with the administration of OEA, which could be salvaged by timely reperfusion. To visualize the ischemic penumbra and its recovery, positron emission tomography (PET) of [ 18 F] uoro-2-deoxy-D-glucose ([ 18 F]FDG) was employed to assess the cerebral glucose metabolism [36,37]. PET data were acquired for 20 minutes at 0 h, 1 h, and 2 h after reperfusion.
Since the in ammation-related high-[ 18 F]FDG uptake might appear around 3 d post-reperfusion [38], PET data were collected within 2 h post-reperfusion to eliminate the effect of in ammation. As shown in Fig.  2a, the right cerebral hemisphere of all the operated rats exhibited signi cantly reduced glucose metabolism compared with the left cerebral hemisphere at 0 h. The reason was that the blood supply of the right cerebral hemisphere was cut off and [ 18 F]FDG could not arrive. Then, with 1 h of reperfusion, the hypoperfused area began to recover in all the operated rats, and their uptake of [ 18 F]FDG began to increase. Interestingly, about 41.7% of the right cerebral hemisphere from the rats administrated with OSDP LNPs exhibited signi cantly elevated signals compared with the normal tissues (the blue dashed outline in Fig. 2a). And the enhancement extended to 83.6% of the right cerebral hemisphere at 2 h postreperfusion. The PET intensity of these regions was about 63.24 ± 2.43 kbp/mL, which was much higher than that of the left cerebral hemisphere (51.76 ± 2.58 kbp/mL). In these regions, the cutting-off blood ow and thereby glucose supply is compensated by an increase of glucose uptake and phosphorylation rate to maintain cellular energy consumption [39], which might be used for repairing the cell damage caused by hypoperfusion. The enhanced glucose metabolism also stated the good cell viability in the region, and the region was called "the ischemic penumbra" in clinic. The ischemic penumbra would recover and functioned well afterwards, which could be veri ed via the TTC-stained brain slices (Fig. 2b). It could be observed that almost all the hypoperfused brain tissues, whose blood supply had been cut off for 90 min, were not infarcted via the administration of the OSDP LNPs. The results indicated that the hypoperfused brain tissues with elevated PET signals after reperfusion were mostly likely to recover. On the contrary, the other MCAO rats did not exhibited enhanced glucose metabolism in the ischemic brain tissues. Although the brains were recovering with reperfusion, the low-level glucose metabolism stated their terrible cell viability at 1 h post-reperfusion. What's worse, an obvious infarction core appeared at 2 h post-reperfusion (The red dashed outline in Fig. 2a). While the other hypoperfused brain tissues possessed similar glucose metabolism to that of the left cerebral hemisphere, they had suffered from irreparable damage owing the acute ischemia and might ultimately die. Many hypoperfused brain tissues with certain PET signals at 2 h post-reperfusion were infarcted as shown in the TTC-stained brain slices of 24 h post-reperfusion (Fig. 2b). Then the cerebral infarct volume and the cerebral edema degree of all the rats were calculated (Fig. 2c-d). With a cerebral infarct volume of about 372.2 ± 26.9 mm 3 and a cerebral edema degree of 13.6 ± 1.0%, the operated rats without treatment had a severe brain damage, which would lead to a serious neurological de cit. With the treatment of free OEA, the cerebral infarct volume of the rats was reduced to 332.2 ± 35.3 mm 3 and the cerebral edema degree to 11.3 ± 1.1%. However, there was no signi cant difference between the OEA and MCAO groups (P > 0.05), which was in according with our previous studies [30]. Nevertheless, the cerebral infarct volume was decreased to 78.2 ± 18.4 mm 3 , and the cerebral edema degree was also decreased to a slight level. The data stated that the rats treated with OSDP LNPs suffered from much less brain damage than the rats of the other groups. What's more, the photos of the TTC-stained brain slices, plus the PET data, forcefully demonstrated that the acute ischemic brain tissues could be preserved as penumbral tissues to a great extent via the administration of OSDP LNPs. And the preserved penumbral tissue would bounce back with reperfusion and possess an increased glucose metabolism within hours to compensate the reduced supply, leading to a better and more rapid recovery.
As an acute disease, ischemic stroke possesses a high mortality, and reducing death rates must be the primary task to a formulation for stroke. As shown in Fig. 2e, only 45.8% of the MCAO rats could survive for 14 d, and the administration of free OEA could not obviously change this data. In comparison, OSDP LNPs could effectively protect from MCAO and increase the survival rate to 83.3%. Furthermore, we speculated that the rats administrated with OSDP LNPs were more likely died from the complications of MCAO, not from the cerebral injury. The small cerebral infarct volume illustrated that they suffered from a little cerebral injury (Fig. S6). The result indicated that the administration of OSDP LNPs might extend the time window for bene cial reperfusion.
Since the behavior of the stroke patients was one of the most important evaluation indicators in clinic, the behavior ability of the rats was evaluated via Garcia method. The Garcia scores of the rats were evaluated at 1, 3, 5, 7, and 14 d after operation. As shown in Fig. 2f, the sham-operated rats got full marks, while the operated rats only got about 10 points in the 1st assessment, illustrating the neurological function de cit of the models and the success of our operation. The rats administrated with OSDP LNPs performed much better than those of the other two operated groups throughout the assessment (p < 0.01). In addition to getting higher scores, the rats administrated with OSDP LNPs maintained enhanced recovery rate within the rst week after operation, which might come from the preserved penumbral tissue. Under the protection of OSDP LNPs, more penumbral tissues were preserved and revived with reperfusion, leading to a better and more rapid recovery. On the contrary, OEA seemed to have no effect on improving neurological function de cit, owing to its extremely low bioavailability.
Morris water maze task.
Learning and memory are one of the most important computational strategies of the brain, which might be impaired by ischemia stroke. To examine the effects of OSDP LNPs on spatial learning and memory, rats were exposed to the water maze task. Between day 15 and 19 after operation, the rats were trained with ve trials per day. Spatial learning ability was assessed by escape latency (the time required to nd the platform in training). Compared with the sham group, the operated rats without treatment exerted much longer escape latency, illustrating their impaired spatial learning ability (Fig. 3A). Owing to their severe brain damage, two rats even could not nd the platform within 120 seconds in the third training, while all the sham rats could reach the platform within 60 seconds in the rst training. Compared with those of the MCAO group, the rats treated with OSDP LNPs needed less time to nd the platform, stating their improved spatial learning ability. Moreover, the training seemed to be more effective to the rats of the nanodrug group, leading to a result that the escape latency was decreased from 75 s in the rst training to 23 s in the fourth training (Fig. 3A). The result also illustrated their strong learning ability. At last, the data had no signi cant difference with that of the sham group, forcefully illustrating that the OSDP LNPs treatment effectively protected the spatial learning ability of the operated rats. Interestingly, the sham-operated rats exhibited signi cantly smarter than the operated rats after the rst training. When entering water in the second training, many of them looked around to con rm the position of the platform, and straightly swam to it. Two rats treated with OSDP LNPs began to have the same performance in the fourth training, while the rats of the other groups just swam around to nd the platform throughout the training.
When the platform was removed, the rats of the sham group and the nanodrug group exhibited markedly increased target crossing times (Fig. 3B), indicating their stronger potential memory for the removed platform. Moreover, it could be seen from the traces that the rats of the sham group and the nanodrug group always swam around the position of the removed platform (Fig. 3C). This strongly indicated that they had signi cant purpose and strong memory for the removed platform. On the contrary, the rats of the other groups possessed much less target crossing times and almost random traces, indicating their little memory for the platform. The results stated that the OSDP LNPs greatly ameliorated ischemia-induced spatial memory impairment.
Immuno uorescence staining and cell counting.
The intractable sequelae of stroke mostly resulted from the injury of neurons, which were really hard to recover. Here, TUNEL staining was employed to evaluate the apoptosis of the neurons in hippocampal CA1, which mainly involved in memory and cognition. As shown in Fig. 3D, none apoptotic cell was detected in the hippocampal CA1 of the sham-operated rats, while a signi cant increase in apoptotic cells could be found in those of the MCAO group and the OEA group. The results stated that the MCAO was likely to induce the apoptosis of the neurons, and this could not be improved by the administration of free OEA. In contrast, only one apoptotic cell was observed in the eld of the nanodrug group. This signi cant decrease was likely related to the improved learning and memory ability of the rats of the nanodrug group. The result also indicated the nice neuroprotective effect of OSDP LNPs.
According to our previous studies [35,40,41], OEA could inhibit the in ammation of reperfusion, which was one of the most primary causes of the brain injury. Microglia and astrocyte play paramount roles in the brain in ammation, which could be speci cally marked by Iba-1 and GFAP, respectively. Hence, the quantity of expressed Iba-1 and GFAP was used to evaluate the in ammation of reperfusion. As depicted in Fig. 4A and Fig. 4B, both markers suggested that the in ammation was slight in the sham-operated rats, while a signi cant increase could be observed in the cortex around the ischemic focus of the MCAO group and OEA group. With the administration of OSDP LNPs, the in ammation was decreased to a low level. The result indicated that the OSDP LNPs could greatly alleviate the in ammation induced by ischemic reperfusion, and therefore provide signi cant neuroprotective effects.

Conclusion
In summary, the study demonstrates a simple but novel method to fabricate internal crystallized liposomes to e ciently encapsulate highly hydrophobic OEA. The hydrogen bond within OEA-SPC totally changed the form of OEA and enhanced its solubility. Then, the crystalization of OEA inside the OSDP LNPs further enhanced the drug loading, and hence, the bioavailability of OEA would highly increase. With further studies focused on the assessment of the neuroprotective effect of the liposomes, the various indicators of the MCAO rats, including the survival rate, the behavioral score, the spatial learning and memory ability, the cerebral infarct volume, the edema degree, the apoptosis of the neurons and the in ammation within the brain, were greatly improved via the administration of OSDP LNPs. What's more, it could be clearly visulized that almost all the acute ischemic brain tissues were preserved under the protection of OSDP LNPs and quickly recovered with reperfusion. These results suggest that the internal crystallized liposomes have potential applications in the eld of hydrophobic drug delivery, and the OSDP LNPs may open new opportunities for the clinic use of hydrophobic OEA as a potential anti-stroke candidate.

Preparation of OSDP LNPs
The OSDP LNPs was synthesized by a nanoprecipitation technique. In brief, OEA (5 mg), and OEA-SPC (3 mg of OEA) were codissolved in 10 mL of DCM, and then the clear homogeneous solution was then dropwise (0.2 mL/min) introduced into distilled water (40 mL) containing DSPE-PEG (20 mg) under magnetic stirring (200 rpm/min). After stirring for one hour, the system turned into a stable, white O/W suspension. Subsequently, the dichloromethane was rapidly removed by rotary evaporation, producing a clear suspension and resulting in the formation of the OSDP LNPs. The OEA-SPC NPs were prepared in the same way except no extra OEA was added in this procedure and DSPE-PEG was replaced by SPC.

Characterization
The OEA − SPC was analyzed using XRD (Phillips X'pert Pro Super), and H 1 NMR (AVANCE III 400 MHz). The bulk OEA powers and SPC were used as control.
The OSDP LNPs were also analyzed with XRD to examine the form of OEA in OSDP LNPs. Morphology of the OSDP LNPs was examined by TEM (Tecnai G2 F20, USA) at 200 kV. The Size and zeta-potential values were determined by a Malvern Zetasizer Nano-ZS machine (Malvern Instruments, Malvern). Three parallel measurements were carried out to determine the average values. The content of OEA in OEA-SPC NPs was determined by LC-MS (LCMS ORBITRAP VELOS PRO ETD, USA). The content e ciency was calculated by Equations (1)

Drugs administration
The OSDP LNPs were dissolved in saline with ultrasonic breaking. OEA and Twain's 80 was dissolved in saline under strong shaking. Drugs (1.5 mg/kg, iv) were administered once in rats at the time of redispersion, and daily for 14 consecutive days after ischemia.
Preparation of the focal cerebral ischemia model Focal cerebral ischemia was induced by middle cerebral artery occlusion (MCAO) in adult male Sprague-Dawley (SD) rats, as previously described. In brief, rats were anesthetized with chloral hydrate (400 mg/kg, ip). The 4 − 0 silicon rubber-coated nylon mono lament was inserted into the right internal carotid artery (ICA) of the rats through the external carotid stump, and past the ECA/ICA bifurcation to occlude the origin of the middle cerebral artery (MCA) at the junction of the circle of Willis. The mono lament was kept in place for 90 min then withdrawn. Sham-operated rats were treated with an identical surgery except that the intraluminal lament was not inserted. Rats were excluded if hemorrhage was found in the brain slices or at the base of the circle of Willis during postmortem examination.

Positron Emission Tomography
PET scans at various time points post-operation (0 h, 1 h, and 2 h), image reconstruction were performed using a nanoScan® PET/CT (Mediso, HUN). MCAO rats (n = 6) were injected with [ 18 F]-2-uoro-2-deoxy-D-glucose (FDG) via the tail vein and 20 min PET scans were performed. Since the best imaging time of FDG was about 45 min after administration, FDG was injected 45 min before each detection time.

Measurement of infarct volume
After PET scans, the rats (n = 6) were decapitated at 24 h post-reperfusion, and the brains were removed rapidly and cut into six 2-mm thick coronal sections, which were then stained with standard 2% TTC at 37°C for 10 min followed by overnight immersion in 10% formalin. Images of the stained brain sections were captured using a digital camera. The infarct area on each TTC-stained section was measured with Image Tool 2.0 software and calculated as the infarct area thickness (2 mm). The summed infarct volume of all brain sections was calculated as the total infarct volume. To determine the extent of ipsilateral oedema, the percentage increase in the ischemic hemisphere volume was calculated by Equations (2):

Measurement of the survival rate
The survival rate of the rats were recorded. (n = 12 rats for sham group and the nanodrug group, and n = 24 rats for the MCAO group and the OEA group) Evaluation of neurological de cit: At 1 d, 3 d, 5d, 7 d, and 14 d after reperfusion, rats (n = 10) were evaluated neurologically by a single examiner who was blinded to the animal groups using the Garcia score. Animals were given a score of 0 to 18 (higher scores indicate greater function).

Morris water maze task
Morris water maze (MWM) task was performed from day 15 to day 20 after reperfusion. The protocol followed the previous report. Acquisition training consisted of 5-days conditioning with ve trails per day from day 15 to 19. For each trail, the rats (n = 8 for each group) were placed in water and allowed to swim for a maximum of 120 s. If the rats found the platform, they were allowed to remain on it for 15 s. If they cannot nd the platform within 120 s, they would be guided to the platform. The escape latency of nding the platform were recorded. On day 20, the platform was removed and rats were given one 120-s retention probe test. We recoded the swimming traces of the rats by a video camera with a computer via an image analyzer. The number of times each animal crossed the position where the platform had been previously located were measured by the analyzer.

Statistical analysis
The statistical signi cance of treatment outcomes was assessed using one-way/two-way analysis of variance for the differences within treatments followed by Tukey's post hoc test (Prism 7 for windows, GraphPad Software Inc., USA); P < 0.05 was considered statistically signi cant in all analyses (95% con dence level). Availability of data and material All data generated or analyzed during this study are included in this published article.