Insight into theranostic nanovesicles prepared by thin lipid hydration and microfluidic method

Liposomes are phospholipid-based self-assembled nanoparticles. Various components can be solubilized in the lipid bilayer, encapsulated in the aqueous core or attached to the surface


Introduction
Liposomes are self-assembled systems composed of phospholipids forming bilayers [1].There are several methods of liposome preparation allowing various scalability potential from the benchtop to the industrial scale.These methods have been classified into three categories: 1) mechanical, 2) based on the replacement of organic solvent(s) by aqueous media, 3) based on detergent removal [2].The most common liposome manufacturing method involves a lipid film formation and subsequent hydration of the film with an aqueous buffer [2].This results in large and heterogeneous particles requiring the addition of a down-sizing technique such as extrusion or sonication [2].The development of ethanol injection and microfluidic techniques has boosted lipid-based nanoparticles' manufacturing using these methods [3][4][5].Mostly, microfluidics has been gaining an increasing interest in liposome production [5][6][7][8].Various microfluidic systems were designed and developed for laboratory use and an industrial scale [9].It was demonstrated that microfluidics is robust, scalable and very reproducible [7,10,11].In this technique, lipids dissolved in an organic solvent are mixed with an aqueous phase through a micromixer.Liposomes of defined sizes can be accurately produced by controlling the microfluidic operating parameters such as flow rate ratio (FRR), total flow rate (TFR), organic solvent and lipid concentration [9,10].
Carugo et al. [9] suggested a gap regarding the exploitation of microfluidics in liposome production compared to standard techniques.In the available literature, only a few examples are addressing the issue [10][11][12].The two studies focused on reaching liposomes prepared by microfluidics and TLH/extrusion provided divergent conclusions [11,12].Noteworthy, both TLH/extrusion and microfluidics were used to produce empty, non-loaded liposomes while drug substances, i.e., vinblastine [11], dexamethasone [12], were loaded by a gradient-assisted method.Kastner et al. [10] showed that vesicles produced with the microfluidics were smaller, had higher drug substance (propofol) loading and a narrower size distribution than those obtained by TLH.It was noteworthy that propofol was loaded during film hydration, and sonication was used as the down-sizing technique replacing extrusion [10].
We have recently demonstrated theranostic liposomes (named GdLip) containing lipid derivative of commercially available MRI CA, Magnevist® and a model photosensitizer zinc phthalocyanine (ZnPc) for photodynamic therapy (PDT), manufactured by TLH.The results have shown that the presence of ZnPc in the theranostic liposomes improved the relaxivity of GdLip [16,27].We now propose a microfluidic method to obtain theranostic liposomes starting with formulations selected in the previous work [16].The presented study aims to evaluate the microfluidic method's applicability for one-step preparation and simultaneous loading of MRI CA and a photosensitizer.Our research is focused on comparing vesicles prepared by TLH and microfluidics at different flow parameters (FRR, TFR).

Liposome preparation and characteristics
Liposomes were prepared by thin lipid film hydration (TLH) followed by extrusion and microfluidic methods (MM), described in detail in the Supplementary material (SM).The amount of ZnPc incorporated into liposomes was determined spectrophotometrically.POPC concentration in liposomes was determined using an enzymatic assay Phospholipids (mti-diagnostics GmbH, Idstein, Germany).Gd's total concentration was determined by a relaxometry measurements at 16.5 MHz after mineralization, following the procedure described by Filippi et al. [28].Size and ζ-potential measurements were performed using DLS instrument (DelsaMax, Denmark ApS c/o, Copenhagen, Denmark).Cryo-transmission electron microscopy (cryo-TEM) and nuclear magnetic resonance (NMR) experiments were carried out following the previously described procedures [16].

In vitro cellular experiments
Squamous cell carcinoma SCC-25 (ATCC® CRL-1628™) and FaDu (ATCC® HTB-43™) cell lines, derived from tongue and pharynx, respectively, were purchased from the American Type Culture Collection (ATCC, Manassas, USA).Details of cell culture, cytotoxicity study and photodynamic assays are provided in the SM.

Loading efficacy
The recovery of lipid components (POPC, PE-DTPAGd) and ZnPc loading efficacy in 16 samples manufactured by TLH/extrusion or MM are presented in Table 1.The highest POPC recovery above 80 % was shown for formulations prepared by M1 (TLH/extrusion; Fig. 1A, B).It varied from ca. 45.9-83.1% in samples obtained by MM and was lower for M3 and M4.The POPC recovery of 72.1-83.7% was shown for samples prepared by the M2 (FRR 3:1, TFR 12 mL min − 1 ).The above results contrast to Kastner et al. [7], who suggested that lipid content remains independent of flow rates and flow ratios.However, a direct comparison is not possible since their samples consisted of different phospholipids and were prepared at lower TFR (0.5 -2 mL min − 1 ) and different FRRs.
PE-DTPAGd loading was correlated (r = 0.9545) with POPC recovery, as shown in Fig. 1C.In our study, PE-DTPAGd was incorporated into the liposomal membrane by dispersing the pure compound in the aqueous solution at 50 • C. On the contrary, Kotouček et al. [23] prepared paramagnetic liposomes by microfluidics while Gd-lipid was dissolved in the organic solvent (ethanol/DMSO mixture) and, as such, injected into the micromixer.The liposomes were used to visualize the morphology of liposomes formed by MM; however, no data about Gd-lipid recovery have been presented [23].
The loading efficacy of the ZnPc photosensitizer was not dependent on the POPC recovery (Fig. 1D) but was highly affected by the preparation method, favoring MM.For TLH/extrusion (M1), 23.8 % and 28.5 % of ZnPc could be loaded into F2M1 and F4M1, respectively.Using MM, we were able to increase ZnPc loading by a factor of 2 in F4M4-42.5 % and almost of 3 in samples M4M2-70.4 %, M4M3-76.2 % compared to vesicles prepared by TLH/extrusion (Fig. 1B, Table 1).Our results are in agreement with the one presented by Kastner et al. [10].They demonstrated three times higher loading of propofol when microfluidics was used in comparison to TLH/sonication.Leaver et al. [21] reported high loading (ca.80 %) of verteporfin by using the MM (TFR 12 mL min − 1 ).We observed that a higher mixing rate (12 mL min − 1 ) resulted in higher loading of ZnPc.Similar observations were reported by Guimarães Sá Correia et al. [29] for quinine loaded at 20 mL min -1 and 6 mL min − 1 .This higher drug loading may result from the highly efficient mixing during microfluidics that favors the hydrophobic molecule incorporation within the bilayers in the same process as the vesicles form [10].

Size and size distribution
The intensity weighted mean hydrodynamic size, z-average (z (av) ) and PDI values of studied liposomes are presented in Fig. 2. Z (av) of all samples prepared by TLH/extrusion (M1) is ~60 nm with PDI < 0.1, depicting a monomodal size distribution.The size of liposomes formed during MM (M2-M4) was 2-3 times smaller than vesicles generated by TLH/extrusion, resulting in vesicles with Ø20− 35 nm (Fig. 2B-D).Zhigaltsev et al. [8] have shown that "limit-size" lipid-based systems can be efficiently produced using a rapid microfluidic mixing technique.
"Limit-size" liposomes are defined as the smallest achievable aggregates compatible with the packing properties of their molecular constituents [8].The different flow parameters used during preparation did not result in smaller particles' formation, indicating that "limit-size" conditions have been achieved.The vesicles with the smallest z (av) , not exceeding 20 nm, were observed in formulations F1.We watched a tendency that the size of liposomes prepared by MM (Fig. 2B-D) increased when ZnPc or PE-DTPAGd was incorporated into POPG/POPC liposomes, resulting in formulations F2 or F3.The size of liposomes F3 was larger than F2.The size of theranostic liposomes F4 prepared by methods M2-M4 was lower than the size of formulations F3 and comparable to vesicles of composition F2.For liposomes prepared by microfluidics (M2-M4),  obtained sizes were not dependent on the flow parameters (Fig. 2B-D).Generally, for these samples, a higher PDI can be observed than for samples resulting from TLH/extrusion.The FDA guidance for industry for liposomal products mentions size and size distribution as critical quality attributes.Still, it does not define the criteria for an acceptable size or PDI [30].Research in the liposome field highlights that efficacy, toxicity and pharmacokinetics of liposomal drug substances can be affected by vesicle size and PDI [11,31,32].Limit size liposomes with particles less than 50 nm accumulated better in the stromal-rich tumors than 100 nm-sized liposomes [8,33].And the general trend is the smaller the particle, the longer circulation time [31].
The intensity-based size distributions derived from a regularization analysis for theranostic samples F4 are shown in the regularization graphs presented in Fig. 3.The regularization analysis produces an estimate of the radii and relative abundance of all species present in the solution without assuming an underlying distribution.
The histogram presented in Fig. 3A corresponds to one particle population, so the F4M1 sample is characterized by monomodal size distribution.The sample's polydispersity is 25.2 %, which is higher than the PDI calculated in cumulant analysis.More than only one peak, linked to specific liposomes' population, can be distinguished in three other histograms received for samples F4M2, F4M3 and F4M4 (Fig. 3B-D).That means these samples are characterized by multimodal size distribution.To be resolved as a separate peak, a species must have a size larger than another species by a factor of five or more and be detectable by the instrument.DelsaMax estimated the relative amount of light scattered based on the several possible particle scattering properties for each mode.The relative amount of mass of a peak of less than 1% can be considered to be negligible.This is valid for sample F4M3 (Fig. 3C), for which the relative mass of peaks 1 and 2 is 99.2 % and 0.8 %, respectively.Therefore the size distribution in sample F4M3 can be considered as monomodal monodispersed.According to regularization analysis, the relative mass of peak 2 in the histograms of sample F4M2 (Fig. 3B) and F4M4 (Fig. 3D) exceeds 1%, therefore in these cases, the distribution should be regarded as bimodal.The measured polydispersity for each peak (peak 1-6.2%, peak 2-11.5%) of F4M2 is low, which means that both species are homogeneous.On the contrary, the particles' distribution in the sample F4M4 is bimodal and is characterized by lower homogeneity because both species' polydispersity is higher than 15 % (peak 1-29.7%,peak 2-22.1%).When the homogeneity level is low (percent polydispersity greater than 30 %), the particle population can be considered to contain significantly different sizes or polydisperse systems.

Short-term stability
Neither size nor PDI of samples prepared by TLH/extrusion (M1) was affected during 60-day storage at 4 • C (Fig. 2A), excluding a slight increase in z (av) of sample F3M1.In the case of samples prepared by M2 (FRR 3:1, TFR 12 mL min − 1 ), a statistically significant difference (p = 0.05) in the particle size of samples F1M2, F2M2, F2M3 and size distribution in F1M2, F2M2, determined on the preparation day and after 60-day storage, can be noticed (Fig. 2B).Furthermore, the size of liposomes prepared by M3 (Fig. 2C) and M4 (Fig. 2D) was mostly unchanged over 60-day storage.Only drops in the z (av) were observed in samples F3M3, F4M3 (Fig. 2C) and F3M4, F4M4 (Fig. 2D).But in the latter samples (F3M4, F4M4) prepared by M4, these differences were not statistically important.The determined ζ-potential of theranostic liposomes (-67.6--62.7 mV) showed that the colloidal system is stable.The reported values were independent of the preparation technique and composition.

Cryo-TEM images
Morphology and size distribution of theranostic liposomes were visualized and analyzed by cryo-TEM (Fig. 4).It is worth underlying statistically, histograms, presenting size distribution of liposomes, are based on a significantly lower number of particles than DLS analyses; however, they provide direct information about their possible shape in colloidal solution.
Cryo-TEM micrographs showed that theranostic liposomes in the studied samples exist mainly in the form of unilamellar vesicles.The percentage of unilamellar liposomes' fraction in the studies samples is about 93 % for F4M1, 95 % for F4M2, 92 % for F4M3 and 85 % for F4M4.
We found that the diameters of liposomes expressed as median values (d med ) and their size distribution obtained from cryo-TEM experiments show relatively close correspondence to DLS measurement results (both resulting from cumulant analysis and intensity-weighted).The median values of liposomes` diameters calculated based on histograms are 51.6 nm (F4M1), 35.2 nm (F4M2), 18.5 nm (F4M3) and 34.0 nm (F4M4) (Fig. 5 A-D).The most evidence discrepancies between data obtained by DLS (z av = 24.3nm) and cryo-TEM (d med = 35.2nm) and the considerably different size distribution character have been noticed in the case of sample F4M2.For this sample, the size distribution calculated from cryo-TEM is broader compared to DLS results.These inequalities between results show the different limitations of the techniques and should as such be considered complementary methods.Notably, sizing liposomes using only one technique may not exactly reflect a complete view of particle size in the studied sample.

Relaxivity analysis
The MRI CA can be encapsulated either in the liposome's inner space [34,35] or as a chelator and lipid hybrid embedded in the bilayer [16].In our studies, we used the latter approach, and PE-DTPAGd was incorporated in the membrane so that gadolinium atoms could interact with protons of water molecules localized outside and inside the vesicles and with water molecules, which constantly permeate the liposomal membrane.
Interestingly, the relaxation properties analysis showed that relaxivities r 1 and r 2 were not dependent on Gd(III) concentration determined in the individual samples (Fig. 5A-D).
The ratio r 2 /r 1 determines whether an agent is most efficient as a T 1 or T 2 ca.For T 1 CA, the r 2 /r 1 ratio should be as close as possible to 1, and if the r 2 /r 1 is >2, the material is predominantly T 2 agent [6,30].At the resonance frequency 400 MHz at 37 • C, the r 2 /r 1 ratios calculated for F3  We compared relaxivities of theranostic formulations F4 (ZnPc/PE-DTPAGd/POPG/POPC) and F3 (PE-DTPAGd/POPG/POPC) prepared by the same method.We observed no differences or only a slight decrease of relaxivity in theranostic liposomes prepared by TLH/extrusion (M1) and microfluidics M2 (FRR 3:1, TFR 12 mL min − 1 ).On the other hand, relaxivities r 1 and r 2 of samples F4, which were obtained via methods M3 (FRR 5:1, TFR 12 mL min − 1 ) and M4 (FRR 3:1, TFR 8 mL min − 1 ), profoundly increased in comparison to values determined for ZnPc-free liposomes F3 (Fig. 5).Like in previous studies, we have observed the positive influence of ZnPc on relaxivities of theranostic liposomes [16,27].We could correlate this impact neither with vesicle size nor concentration of Gd ions and ZnPc embedded in the vesicles.We have performed the modified Florence model fitting analysis, which has shown that the change in the relaxivity parameters of theranostic liposomes upon ZnPc incorporation is not associated with the appearance of an additional paramagnetic center [27].The higher relaxivities of Gd (III) ions in the theranostic liposomes compared to ZnPc-free paramagnetic liposomes PE-DTPAGd/POPG/POPC can be attributed to the changes that occur inside the liposomal bilayer and affect water permeability across the liposomal membrane.In the literature, the permeability of water across the membrane is usually explained by the solubility-diffusion model [38].It is well-recognized from cell biology that increasing the surface area (SA) to volume (V) ratio makes diffusion or osmosis much more effective.We calculated the SA/V ratio using cryo-TEM analysis data and found that relaxivities increased proportionally to the SA/V ratio (Fig. 6A,B).
Also, we noticed that r 1 and r 2 parameters' values decrease with increasing POPC/ZnPc molar ratio in an exponential manner (Fig. 6C,  D).It is known that a hydrophobic component placed in the vicinity of phospholipid chains may lead to the stiffening of the unsaturated phospholipid chains [29,30].We hypothesize that increased content of ZnPc incorporated into theranostic liposomes may cause stiffening of the structure in one place of the liposomal membrane.The stiffening may facilitate the formation of membrane defects in areas where only phospholipids are present, resulting in an increase in diffusion through the membrane and increased participation in the relaxation of magnetic centers inside liposomes [27,29,30].
Another aspect influencing the relaxation properties is the preservation of the lipid membrane surface.For a system in which the bilayer consists of only unsaturated phospholipids (such as POPC), the membrane's surface may fluctuate, making it difficult for water molecules to access paramagnetic centers located on it [39,40,41].However, simultaneously it facilitates local rotational movements of paramagnetic centers, significantly affecting the relaxation properties [40].
In summary, the preparation method indirectly impacts the relaxivity of theranostic liposomes by influencing the size of the liposomes and SA/V ratio as well as components` incorporation efficacy that determines the final POPC/ZnPc molar ratio.

In vitro cellular study
Head and neck cancer is one of the main indications for PDT [42,43].As Cerrati et al. [44] presented in the meta-analysis, PDT has been considered effective as a primary surgical resection treatment for early-stage squamous cell carcinoma of the oral cavity, a good function-preserving approach to treatment.PDT is also potent in the overall management of oral premalignant lesions.MR imaging is a crucial tool in the radiological evaluation of HNSCC.Clinically used PDT preceded and followed by MR imaging is the desired combination enabling treatment of HNSCC and monitoring therapeutic outcomes [43].
The cells were treated for 24 h with ZnPc (0.05-3.80 μM), either free or loaded into theranostic liposomes in the studies.Tested formulations demonstrated limited toxicity in Fadu and SCC-25 when no light (0 J/ cm 2 ) was applied, as shown in Fig. 7A, C.Under dark conditions, cell viability (CV) was reduced to ca. 80 % in SCC-25 cells and ca.60 % in FaDu ones.We observed cytotoxicity neither for the liposome control nor for the free ZnPc in both cell lines.Therefore, within the studied ZnPc concentration ranges, the half-maximal inhibitory concentration (IC 50 ) for dark conditions was not calculated.
The free ZnPc did not show any photocytotoxicity at the studied concentrations and irradiation conditions (Fig. 7B, D).In contrast, Cheung et al. [45] showed a decrease in the CV of FaDu cells upon photodynamic treatment with free ZnPc at 0.5 μM (CV = 45 %) and 1 μM (CV = 10 %).Noteworthy, Cheung et al. [45] used a different light source and two times higher light dose of 3.6 J/cm 2 compared to our set-up.Additionally, Cheung et al. used 0.5 % DMSO in the cell culture medium.We suppose that the decreased activity of free ZnPc observed in our study might result from a lower DMSO concentration in the medium, e.i.0.1 % vs. 0.5 %, which was used by Cheung et al. [45].ZnPc, as a highly insoluble photosensitizer, is pre-dissolved in DMSO and subsequently diluted in a culture medium to reach the final concentration used to treat cells.A decreased DMSO concentration enhances the formation of ZnPc aggregates, which is well-known to attenuate photosensitizers' optical and biological properties [46].
The pronounced photocytotoxic activity of ZnPc in theranostic liposomes compared to free ZnPc was observed following irradiation with a light dose of 1.8 J/cm 2 (Fig. 7B, D).These results indicate that the theranostic delivery system possesses highly favorable performance (irrespectively of preparation technique) since it does not cause any toxicity by itself but deliver ZnPc in a highly potent monomeric

Table 1
Composition, molar ratio, recovery of POPC and PE-DTPAGd, and encapsulation efficacy (EE) of ZnPc in the studied liposomes.