Toward Understanding Polymer Micelle Stability: Density Ultracentrifugation Offers Insight into Polymer Micelle Stability in Human Fluids

Abstract

Micelles, as a class of drug delivery systems, are underrepresented among United States Food and Drug Administration approved drugs. A lack of clinical translation of these systems may be due to, in part, to a lack of understanding of micelle interactions with biologic fluids following injection. Despite the limited clinical translation, micelles remain an active area of research focus and pre-clinical development. The goal of the present study was to examine the stability of amphiphilic block copolymer micelles in biologic fluids to identify the properties and components of biologic fluids that influence micelle stability. Micelle stability, measured via Förster resonance energy transfer-based fluorescent spectrometry, was complemented with density ultracentrifugation to reveal the colocalized, or dissociated, state of the dye cargo after exposure to human biologic fluids. Polymeric micelles composed of poly(ethylene glycol-block-caprolactone) (mPEG-CL) and poly(ethylene glycol-block-lactide) (mPEG-LA) were unstable in fetal bovine serum, human serum and synovial fluid, with varying levels of instability observed in ascites and pleural fluid. All polymeric micelles exhibited stability in cerebrospinal fluid, highlighting the potential for local cerebro-spinal administration of micelles. Interestingly, mPEG_2.2k-CL_3.1k and mPEG_2k-LA₂._7k micelles favored dissolution whereas mPEG_5.4k-LA_28.5k micelles favored stability. Taken together, our data offers both quantitative and qualitative evidence for micelle stability within human biologic fluids and offers evidence of polymer micelle instability in biologic fluids that is not explained by either total protein content or total unsaturated lipid content. The results help to identify potential sites for local delivery where stability is maintained.

Graphical Abstract

Introduction

In recent years, micelles have translated to the clinic with the approval of Genexol-PM, also known as Cynviloq, and Paclical, also known as Apealea. Micelles are widely investigated delivery stystems due to their favorable energy for self assembly and their ability to solubilize hydrophobic drugs [1]. Many amphiphillic block copolymer micelles are composed of synthetic polymers individually used in drug delivery systems approved by the Food and Drug Administration, including polyethylene glycol (PEG), polycaprolactone (PCL), polylactide-co-glycolide (PLGA), and polylactide (PLA). As the hydrophillic component of many micelle-forming block copolymers, PEG has been generally used to improve pharmacokinetics of drug delivery systems based on the theory that PEG diminishes protein and cellular interactions when present at sufficient density [2–4]. PCL, PLGA, and PLA promote the formation of micelles above a critical micelle concentration (CMC) due to favorable thermodynamic energy associated with hydrophobic block aggregation and solvation of PEG [5, 6]. The compatibility of the cargo can further influence the loading, formation, and stability of the micelles [7–9].

Once formed, micelles are believed to transport and deliver a concentrated therapeutic dose to diseased tissues [10] with the potential for diseased-tissue accumulation by passive accumulation [11, 12]. However, there is growing evidence for the limits of passive and active targeting [13, 14]. In order to achieve passive targeting, micelles must be structurally stable enough to maintain their payload in an intact micelle until reaching the diseased tissue. To understand the structural stability of micelles in vitro and in vivo, two primary methods have been utilized, pyrene fluorescence shift [15, 16] and Förster resonance energy transfer (FRET), typically between 3,3-dioctadecyloxacarbocyanine perchlorate (DiO) and 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate (DiI) dyes [17–19]. Even with these methods, there is no clear consensus of the stability of micelles after injection. These results are also complicated by the differentiation of micelle disassembly from cargo release. Cargo release has been demonstrated to be related to micelle core-cargo compatibility [7–9]. For the purposes of this manuscript, instability is describe to include both disassembly of the polymer micelles and premature release of cargo; however, these are two distinct mechanisms of instability.

Clinical efficacy of micelles has underperformed preclinical expectations due to immunogenicity, interactions with other soluble fluid molecules, and stability [20, 21]. Proteins adsorb onto nanoparticles and influence cellular recognition in both a charge and size dependent manner [22–25]. Adsorption of proteins to particles has been shown to ultimately influence particle clearance from the systemic circulation [26]. While little is known about the formation of the protein corona on micelles or the proteins’ influence on stability, there is clear evidence protein coronas form on polymeric materials.

Along with protein interactions, lipid rich particles such as lipoproteins are also witnessed to negatively alter micelle stability [27]. Incubation of mPEG-CL micelles in whole plasma disrupts micelles, with radioactively labeled carrier and drug distributing among several LDL, HDL, and lipoprotein fractions [28]. Lipophilic interactions with micelles further counfound our understanding of micelle stability in physiologic application. The biologic milieu experienced by micelles upon administration has a profound impact on their integrity.

Although much has been learned, there is still little known about the components of biologic fluids that impact micelle stability. Identification of biologic fluids and environments with little detrimental effect on micelle stability would be expected to aid in the understanding of micelle stability by allowing comparision of biologic fluid composition. Little information exists describing micellar interactions with the biological fluids beyond blood [24], and further information on the interactions with relevant biologic fluids is needed [29]. The sparsity of data on the interaction of micelles with complex fluids is most likely due to the difficulty of separating micelles from the bulk solution. Solid nanoparticles may be pelleted via centrifugation while micelles do not readily sediment from the bulk fluid. To address this issue, density ultracentrifugation was used to examine micelle populations within complex fluids. Density ultracentrifugation has routinely been employed both in preparative separation as well as analytical analysis of biologic interactions [30, 31]. Ultracentrifugation techniques have been applied to acellular biologic macromolecules [30], synthetic nanoparticles [32], and cellular components, specifically extracellular vesicles and organelles [33, 34]. In this work, ultracentrifugal separation is combined with FRET spectrometry and other micelle characterization techniques to better understand micelle stability.

Based on our previous observations that micelles exhibited higher stability in specific biologic fluids [35], the goal of the present study was to identify biologic fluids that impacted micelle stability. Specifically, the stability of three micellar block copolymers were examined by fluorescent spectrometry and sucrose gradient ultracentrifugation across a panel of human physiologic fluids at physiological temperature, including ascites (AF), cerebrospinal fluid (CSF), pleural fluid (PF), synovial fluid (SF), human serum (HS), and fetal bovine serum (FBS), in order to determine if protein and lipid content influence micelle instability within the biologic fluid and to identify administration routes that favor micelle stability in human fluids.

Materials & Methods

Micelle preparation and characterization

A mass of the block copolymer, mPEG_2.2k-CL_3.1k (Polyscitech; AK73; 403ASMS-A), mPEG_2k-LA_2.7k (Polyscitech; AK009; 180329YSK-A), or mPEG_5.4k-LA_28.5k (Polyscitech; AK54; 50320ZZZ), was dissolved in acetone (3 mL). For each, the subscript describes the number average molecular weight for the respective block of the copolymer. Fluorsecent dyes, DiO (Marker Gene Technologies; M1197; 216KJL145) and DiI (Life Technologies; D282; 1751738), were dissolved in a 50% acetonitrile and 50% methanol mixture. An appropriate amount of dye solution to achieve 5 μg of each dye per 1 mg of polymer, or a final 10 μg total dye, was stirred into the acetone polymer solution in a parafilm-covered borosilicate glass test tube for 30 mins before dropwise addition of water (1 mL). This solution was allowed to stir for 30 mins before being dialyzed against deionized water in a wetted regenerated cellulose ester dialysis membrane (3.5 kDa molecular weight cutoff; Spectrapor Labs; 132724) for 6 changes of dialysate over at least 24 h. Micelles were then syringe filtered through a 0.22 μm PVDF membrane filter (Millex-GV; SLGVM33RS) and characterized via dynamic light scattering (DLS).

Dye free micelles were prepared similarly with only the omission of dyes. To determine polymer concentrations, the polymer mass recovered after dialysis of micelles was determined following freeze drying (Freezone 4.5; Labconco) of a 1 mL aliquot. In this way, standard curves for the polymer in water were generated (Supplementary Figure 1) in a quartz cuvette on a DU800 UV-Vis spectrometer (Beckman Coulter) [36, 37]. The concentration of the polymers in all experiments were then determined from these micelle standard curves.

Dynamic light scattering was performed using 50 μL of filtered micelle solution in disposable microcuvettes (Fisher Sci, NC0628994) utilizing a Zetasizer Nano ZS (Malvern Inc) at room temperature. Micelle diameter (nm) was determined via cumulant fit analysis (Malvern Inc). Zeta Potential was performed using 1 mL of filtered micelle solution in 500 μM HEPES, 20 μM NaCl at pH 7.0 in a disposable folded capillary cell (Fisher Sci, NCO491866) utilizing a Zetasizer Nano ZS (Malvern Inc) at room temperature. Zeta potential (mV) was determined via phase analysis light scattering under monomodal analysis at 40V.

Critical Micelle Concentration (CMC)

Serial dilutions of micelle solutions (2 mL) were created in amber vials to which pyrene stock (5 μL; 800 μM in acetone) was added to a final concentration of 2 μM pyrene. Micelles were stirred overnight, covered from light, to allow for maximal evaporation of acetone and partition of dye into micelles. Fluorescence was measured at an excitation wavelength of 333 nm and emission wavelengths of 373 nm (λ_I) and 388 nm (λ_II). The ratio of the fluorescent intensity (intensity ratio) was calculated by dividing the fluorescent intensity at λ_II by the fluorescent intensity at λ_I. The inflection point, determined by plotting the best fit lines and calculating the intersecting point, of the polymer concentration versus intentity ratio was used to determine CMC [15].

Protein and Lipid Concentration Assays

Bicinchaninnic acid (BCA) protein assay was performed with 3 separate aliquots of each fluid as described by the manufacturer using a Pierce BCA Assay Kit (Thermo Fisher Scientific Inc.). The sulfo-phospho-vanillin assay was used to measure the total unsaturated lipid content in biologic fluids per the manufacturer’s 96 well plate protocol (Cell BioLabs, Inc).

Fluid-Micelle Interaction

Dye loaded micelle stocks were diluted into 1X PBS (Corning; 21–013CV). Ascites fluid (AF; 991–07; lot #: 05B4970), cerebrospinal fluid (CSF; 991–19-P; lot #: 03C3076), pleural fluid (PF; 991–38; lot #: W128509), human serum (HS; 991–24-P-PD; lot #: 03C2266) and synovial fluid (SF; 991–42-P; lot #: 12B2882) were obtained from human subjects by Lee Biosystems under institutional review. All human fluids except for ascites and pleural fluid were obtained from normal donors. PBS, human fluids, or FBS (Gemini Bio; 900–108; lot #: A25F00H) were added to the sample and the solutions were incubated in Eppendorf tubes (covered from light) in a waterbath at 37°C for density ultracentrifugation experiments or in disposable semimicro methacrylate cuvettes for FRET spectrometry assays.

FRET Spectrometry

Micelle stocks were diluted with PBS and the corresponding biologic fluid in a disposable cuvette to a total volume of 500 μL. Fluorescence emission was measured on a RF 1501 spectrofluorimeter (Shimadzu Scientific) in disposable semimicro methacrylate cuvettes (Fisher Scientific;14-955-128). The micelle fluid mixture was mixed via pipette and the solution was excited at a wavelength of 480 nm. Wavelength emission was scanned upon excitation at 480 nm from 490 nm to 600 nm where the intensity of the acceptor (DiI) at a wavelength of 565 nm (I_DiI) and the emission intensity of donor (DiO) at a wavelength of 502 nm (I_DiI) was measured. Cuvettes were covered from light and incubated for 2 days at 37°C in a waterbath with wavelength scans at various times. The FRET ratio was calculated as the ratio of fluorescence intensity at the acceptor wavelength, I_DiI, divided by the sum of the fluorescence intensity at acceptor and donor wavelenths (I_DiI + I_DiO).

Sucrose Gradient Ultracentrifugation

Sucrose (Sigma Aldrich; S9378) was dissolved into de-ionized water to 60% w/v by magnetic stirring overnight and diluted as needed with distilled, deionized water. Gradient layers were created in ultracentrifuge tubes (Beckman Coulter Inc.; 331372) by sequential freezing on dry ice of 1 mL layers at decreasing 5% sucrose intervals from 60% to 20%. Frozen gradients were thawed for 8 h at 4°C and placed at room temperature 30 min prior to sample layering. While gradients thawed, FRET micelles (1 mg/mL) were mixed with 0% (PBS), 10%, or 90% biologic fluid at appropriate times prior to being placed in a 37°C waterbath and covered from light. The 90% biologic fluid sample is considered the most concentrated sample that could be created while accounting for dilution with micelle solutions. Samples were loaded on sucrose gradients via a 1 mL syringe with no needle. Sucrose gradient density ultracentrifugation separation was performed in a LE-80 ultracentrifuge (Beckman Coulter) with a SW41 TI swinging bucket rotor (Beckman Coulter) at 247,000x g for 24 h at 20°C. Sucrose gradients were visualized on a UV Transilluminator (Fotodyne 3–3000) and photographed with a digital camera.

Statistical Analyses

Two-way ANOVA was used to test FRET timepoint significance while one-way ANOVA was used to test lipid and protein concentrations. Post-hoc Tukey analysis was utilized if ANOVA suggested significant differences between the groups with significant p-values less than 0.05. All quantitative experiments were independently replicated at least three times and presented as means plus or minus the standard error.

Results

Characterization of Block CopolymerMicelles

All block copolymers exhibited the expected discontinuity in pyrene fluorescence indicative of micelle formation (Supplementary Figure 2). The CMCs of mPEG_2.2k-CL_3.1k, mPEG_2k-LA_2.7k, and mPEG_5.4k-LA_28.5k were 4.9±1.8 μg/mL, 1.1±0.2 μg/mL and 0.5±0.2 μg/mL, respectively.

All micelles exhibited a monomodal distribution, but the mPEG_2k-LA_2.7k micelles exhibited a larger diameter (72.9±26.4 nm) than the mPEG_2.2k-CL_3.1k (27.3±11.1 nm) and mPEG_2k-LA_2.7k (27.1±10.0 nm) micelles (Supplementary Figure 3). This was expected as the mPEG_5.4k-LA_28.5k polymer number average molecular weight (M_N) was 34,070 g/mol with an approximately 29,000 g/mol poly(D,L-lactide) block compared to mPEG-CL with a M_N of 5,426 g/mol and an approximately 3,200 g/mol polycaprolactone block and mPEG_2k-LA_2.7k with a MN of 4,729 g/mol and an approximately 2,778 g/mol poly(D,L-lactide) block [38, 39]. The zeta potential of mPEG_2.2k-CL_3.1k, mPEG_2k-LA_2.7k, and mPEG_5.4k-LA_28.5k micelles was −6.7±0.4 mV, −7.8±0.8 mV and −14.3±0.4 mV, respectively.

Total protein and unsaturated lipid in biologic fluids

As a base for comparison, the protein content of each fluid was measured (Figure 1A). Synovial fluid contained the highest protein followed by FBS, HS, and pleural fluids. It should be noted that there is no statistical difference between the protein content in these fluids. Ascites fluid had significantly lower protein with CSF having significantly lower protein content than the other fluids. Statistical comparisons are presented in Supplemental Table 1.

Figure 1.

Characterization of (A) protein and (B) unsaturated lipid content in biologic fluids. Bicinchoninic acid and sulfo-phospho-vanillin assays were used to quantify protein and unsaturated lipid, respectively, in human cerebrospinal fluid (CSF), human ascites fluid (AF), human serum (HS), fetal bovine serum (FBS), human pleural fluid (PF), and human synovial fluid (SF). Statistical significance is presented in Supplementary Tables 1 and 2. Bars represent the mean plus or minus (±) the standard error of the mean of three independent experiments.

Similarly, the unsaturated lipid content of each fluid was measured (Figure 1B). Cerebrospinal fluid had the least lipids. Synovial fluid had the most lipid followed by pleural fluid, FBS, HS, and ascites fluid. Statistical comparisons are presented in Supplemental Table 2.

Micelle FRET Change in Human Fluids

When diluted to 200 times the CMC for the given block copolymer in PBS or PBS supplemented with 10% v/v of each biologic fluid, micelles composed of mPEG_2.2k-CL_3.1k and mPEG_5.4k-LA_28.5k showed negligible decrease in FRET intensity (Figure 2A,B). However, mPEG_2k-LA_2.7k exhibited decreasing FRET when mixed with the diluted biologic fluids (Figure 2C). The dissociation of FRET dyes under typical cell culture conditions (10%) was a material specific phenomenon as these results did not match those seen with mPEG_2.2k-CL_3.1k micelles which shared similar HLB, zeta potential, and diameter, but different CMCs (Table 1). Experiments were conducted well above the CMC (200x), so dilution alone does not explain this phenomenon. The mPEG_2k-LA_2.7k micelles exhibited the greatest loss of FRET at 1h in both HS and FBS, followed by synovial fluid, ascites, and pleural fluids while no FRET loss was exhibited in PBS and CSF. At 12h, this trend remained, though the magnitude of FRET loss increased dramatically for HS and FBS compared to synovial, ascites, and pleural fluids.

Figure 2.

FRET intensity is stable following dilution of (A) mPEG_2.2k-CL_3.1k and (B) mPEG_5.4k-LA_28.5k micelles in 10% biological fluid compared to (C) mPEG_2k-LA_2.7k micelles. Micelles exhibit changes in FRET when diluted to 200x the CMC in phosphate buffered saline (PBS; ) or biologic fluids (10%): human cerebrospinal fluid (CSF; ■), human ascites fluid (AF; ▲), human serum (HS; ●), fetal bovine serum (FBS; ◆), human pleural fluid (PF; ▼), and human synovial fluid (SF; ). Points represent the mean plus or minus (±) the standard error of the mean of three independent experiments.

Table 1.

Block copolymer and polymer micelle parameters.

Diblock copolymer (A-b-B)	MN (g/mol)†		HLB‡	micelle diameter (nm)§	ζ-potential (mV)	CMC (μg/mL)§	CMC (M)
mPEG2.2k-CL3.1k	2,200	3,100	8.3	27.3 ± 11.1	−6.7±0.4	4.9±1.8	9.3±3.4×10−7
mPEG2k-LA2.7k	2,000	2,700	8.5	27.1 ± 10.0	−7.8±0.8	1.1±0.2	2.4±0.3×10−7
mPEG5.4k-LA28.5k	5,400	28,500	3.2	72.9 ± 26.4	−14.3±0.4	0.5±0.2	1.5±0.5×10−8

^†Data supplied by supplier.

^‡Calculated from molecular weights.

^§Data calculated and presented as mean plus or minus the standard error of the mean (n = 3).

When mPEG_2.2k-CL_3.1k micelles were diluted in media that was composed primarily of the biologic fluids, the observations were significantly different. The mPEG_2.2k-CL_3.1k micelles exhibited FRET intensity decay within 90% biologic fluid (Figure 3A). The mPEG_2.2k-CL_3.1k micelles exhibited the greatest FRET loss at 1h in SF, FBS, and HS while the FRET observed in pleural fluid, ascites fluid, and CSF remained relatively stable over time.

Figure 3.

FRET intensity following dilution to 90% biologic fluids for (A) mPEG_2.2k-CL_3.1k, (B) mPEG_5.4k-LA_28.5k micelles, and (C) mPEG_2k-LA_2.7k micelles. Micelles exhibit changes in FRET when diluted to 200x the CMC in in phosphate buffered saline (PBS; ) or biologic fluids (90%): human cerebrospinal fluid (CSF; ■), human ascites fluid (AF; ▲), human serum (HS; ●), fetal bovine serum (FBS; ◆), human pleural fluid (PF; ▼), and human synovial fluid (SF; ). Points represent the mean plus or minus (±) the standard error of the mean of three independent experiments.

The stability of the mPEG_5.4k-LA_28.5k micelles was also reduced in higher contents of biologic fluid, though this effect appeared to occur upon initial mixing and remained stable over time (Figure 3B). Synovial and ascites fluid caused the largest decreases in FRET intensity upon mixing and steadily increased afterward. The mPEG_5.4k-LA_28.5k micelles FBS exhibited a decreasing FRET intensity up to 12h and had the lowest FRET intensity of all fluids. In HS and PF, the FRET intensity remained stable upon incubation after mixing with mPEG_5.4k-LA_28.5k micelles. The mPEG_5.4k-LA_28.5k micelles exhibited a less pronounced change in FRET intensity upon dilution in biologic fluids, but the effect is more significant in the higher biologic fluid content than in lower biologic fluid content conditions.

Micelles composed of mPEG_2k-LA_2.7k exhibited much greater loss in FRET induced DiI fluorescence at these biologic fluid concentrations (Figure 3C). This loss in FRET intensity was also significantly greater than observed at lower content (10%) of biologic fluids (Figure 2C). At 1h, FBS exhibited the greatest FRET loss for mPEG_2k-LA_2.7k micelles followed by HS, synovial, and ascites fluid. Pleural fluid and CSF were the least destabilizing fluids at 1h for mPEG_2k-LA_2.7k micelles. The FRET intensity after incubation of all micelles in CSF resulted in an increase that was dependent on the specific diblock copolymer (Figure 3).

After 4h incubation in ascites fluid, mPEG_2k-LA_2.7k micelle FRET intensity increased paradoxically (Figure 3C). These confounding results suggested that further information was needed to interpret the FRET intensity data.

Sucrose gradient visualization of stability

To better interpret the results obtained using FRET, density gradient ultracentrufigation was used to separate intact micelles from the biologic fluids. When FRET dyes were loaded in the micelles, micelles incubated in PBS appeared as yellow bands (Figure 4). The mPEG_2.2k-CL_3.1k micelles had a density that results in a compact fraction between the 20% and 25% sucrose layers whereas mPEG_2k-LA_2.7k micelles sedimented into the 35% sucrose fraction indicating a much more dense micelle. The mPEG_5.4k-LA_28.5k micelles were most dense, resulting in a diffuse fraction at approximately 45% sucrose. As a guide, a black bar has been inset to compare the sedimentation of micelles not incubated with biologic fluid (Figure 4) with the sedimentation observed following incubation with the biologic fluids (Figures 5–10).

Figure 4.

Representative sucrose gradient ultracentrifugal separation visualization of mPEG_2.2k-CL_3.1k (left), mPEG_2k-LA_2.7k (middle), and mPEG_5.4k-LA_28.5k (right) micelles, loaded with DiO and DiI. Micelles incubated in PBS for 6 h at 37°C appeared as stable yellow bands in sucrose gradients at the appropriate density following ultracentrifugation. The sucrose gradient was from 20% (top) to 60% (bottom) sucrose. The black bar behind each image demonstrates the approximate location of the polymer micelle band, this black bar is repeated in other figures as approximate reference to aid the reader.

Figure 5.

Representative sucrose gradient ultracentrifugal separation visualization of mPEG_2.2k-CL_3.1k (left), mPEG_2k-LA_2.7k (middle), and mPEG_5.4k-LA_28.5k (right) micelles following incubation in FBS (10 and 90% diluted with PBS) for varying amounts of time (1 h or 6 h) at 37°C. The black bar behind each image demonstrates the approximate location of the polymer micelle band when no protein is present. The arrow indicates the approximate location of new bands appearing following incubation.

Figure 10.

Representative sucrose gradient ultracentrifugal separation visualization of (left) mPEG_2.2k-CL_3.1k , (middle) mPEG_2k-LA_2.7k , and (right) mPEG_5.4k-LA_28.5k micelles following incubation in cerebralspinal fluid (CSF; 10, 50, or 90% with the remainder PBS) for varying amounts of time (1 h or 6 h). The black bar behind each image demonstrates the approximate location of the polymer micelle band when no protein is present.

When the micelles are incubated with biologic fluids, the micelles are stable for some time, but the micelles’ FRET cargo can be clearly observed to dissociate. As dissociation occurs, the fluorescent dyes were observed outside the range of density for the micelles. Micelle incubation with FBS resulted in the greatest disruption of the yellow fluorescent band and the concomitant appearance of new fluorescent bands which was most pronounced in the 90% FBS samples (Figures 5). This disruption increased with incubation time and as content of FBS increased. A transfer of fluorescence to less dense fractions within the gradients, outside the corresponding micelle fraction, highlight the dissociation of the micelle cargo within FBS.

Individual green-blue bands appear following incubation of micelles with most fluids for the higher fluid content samples. These bands correspond to protein autofluorescence (Supplementary Figure 4) and appeared due to the increase in protein content as biologic fluid concentration was increased from 10% to 90% total sample volume.

Synovial fluid (Figure 6) caused a dramatic decrease in mPEG_2.2k-CL_3.1k micelle band intensity after incubation with 90% synovial fluid within 1 h of incubation. This was less dramatic for mPEG_2k-LA_2.7k micelles, but the intensity of the micelle band noticeably decreased corresponding with the appearance of new fluorescent bands near the 25 to 30% gradient layers. Similar results were observed for both mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles after incubation in ascites fluid (Figure 7) and pleural fluid (Figure 8).

Figure 6.

Representative sucrose gradient ultracentrifugal separation visualization of mPEG_2.2k-CL_3.1k (left), mPEG_2k-LA_2.7k (middle), and mPEG_5.4k-LA_28.5k (right) micelles following incubation in synovial fluid (10 and 90% diluted with PBS) for varying amounts of time (1 h or 6 h). The black bar behind each image demonstrates the approximate location of the polymer micelle band when no protein is present. The arrows indicate the approximate location of new bands appearing following incubation.

Figure 7.

Representative sucrose gradient ultracentrifugal separation visualization of mPEG_2.2k-CL_3.1k (left), mPEG_2k-LA_2.7k (middle), and mPEG_5.4k-LA_28.5k (right) micelles following incubation in ascites fluid (10 and 90% with the remainder PBS) for varying amounts of time (1 h or 6 h). The black bar behind each image demonstrates the approximate location of the polymer micelle band when no protein is present. The arrows indicate the approximate location of new bands appearing following incubation.

Figure 8.

Representative sucrose gradient ultracentrifugal separation visualization of mPEG_2.2k-CL_3.1k (left), mPEG_2k-LA_2.7k (middle), and mPEG_5.4k-LA_28.5k (right) micelles following incubation in pleural fluid (PF; 10 and 90%) for varying amounts of time (1 h or 6 h). The black bar behind each image demonstrates the approximate location of the polymer micelle band when no protein is present. The arrows indicate the approximate location of new bands appearing following incubation.

However, the fluorescent mPEG_5.4k-LA_28.5k micelle band shifted toward more dense gradients corresponding both with increasing biologic fluid volume and time (Figures 7, 8, 9, & 10). This result was in stark contrast to that witnessed in the gradients containing mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles.

Figure 9.

Representative sucrose gradient ultracentrifugal separation visualization of mPEG_2.2k-CL_3.1k (left), mPEG_2k-LA_2.7k (middle), and mPEG_5.4k-LA_28.5k (right) micelles following incubation in human serum (HS; 10 and 90% diluted with PBS) for varying amounts of time (1 h or 6 h). The black bar behind each image demonstrates the approximate location of the polymer micelle band when no protein is present. The arrows indicate the approximate location of new bands appearing following incubation.

The mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelle exhibited similar instability within HS (Figure 9) compared to both FBS (Figure 6) and ascites fluids (Figure 7). The mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles incubated in HS exhibited a dramatic decrease in micelle band intensity, also similar to incubation in FBS. However, mPEG_5.4k-LA_28.5k micelles resembled those results witnessed after incubation in FBS and pleural fluids wherein the micelle band sedimented into much greater density gradient layers.

Finally, micelles incubated in CSF exhibited no visual decrease in micelle fluorescent bands (Figure 10). There also was no appearance of individual dye bands. These results indicats a negligible effect on micelle stability when mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles are incubated with CSF. However, mPEG_5.4k-LA_28.5k micelles appeared to sediment into more dense layers corresponding both to time of incubation and fluid content.

Discussion

The critical micelle concentration describes the concentration where the standard free energy of micellization, the interaction energy between polymer chains, and the interaction energy with the bulk solvent and any solvent components are balanced, i.e. the concentration where the energy driving assembly is greater than the solvating energy. As chain length of the hydrophobic tail increases, the Gibbs free energy of micellization decreases and micelles form at lower concentrations (CMCs). Below the CMC, as will be expected at some point after administration, the micelles would be expected to spontaneously disassemble if the micelles are thermodynamically unstable and kinetically unstable, but this may be a slow process if the micelles are kinetically stable and thermodynamically unstable [40–42]. Micellar association and dissassembly has been demonstrated in lipid and block copolymer micelles [5, 39, 43, 44]. The chain association can also be influenced by the composition of the media. As a thermodynamic parameter determined in well-defined media, a material’s CMC does not consider interactions with ions, proteins, lipids, or other materials present in biologic fluids. Therefore, it is necessary to understand what other parameters may influence stability outside these well-defined, non-physiologic fluids, specifically PBS or water. In order to understand the influence of biologic environments while avoiding dissolution due to dilution, micelle stability was examined at two hundred times the experimentally determined CMC for each copolymer. By working significantly above the CMC in controlled conditions that mimic biologic fluids, the thermodynamic drive for micelle formation is expected to be maintained.

These conditions do not discount the significance of the rapid dilution experienced by micelles upon administration nor the contribution of shear stress during transit. Shear stress has been shown to contribute toward micelle instability in vitro [45]. Many factors may affect this shear stress, such as the viscosity of the fluid, its velocity along the surface, and the distance to the surface where shear stress is experienced. The frictional shear experienced by micelles is therefore subject to change, based not only on the fluid’s physiochemical properties (such as viscosity), but also the architecture of the vasculature and the rate of the fluid’s flow. These additional factors will influence stability of micelle drugs but were beyond the scope of this study.

To understand the assembly and disassembly of micelles in complex media, FRET based assays, typically using DiO and DiI, have been used to observe micelle stability via fluorescence spectrometry [17, 18, 36, 46], but our results shows that this analysis may be overly simplified, e.g. Figure 2 and Figure 3. Both dyes, although not matched to the core chemical properties [7–9], are hydrophobic with LogP values well above 10 and limited solubility in aqueous media. Drug hydrophobicity, polymer and drug miscibility, and heat of mixing all influence drug loading of micelles [47] where the loaded drug’s hydrophobic strength is also believed to increase the micelle’s stability. Therefore, the DiO DiI FRET pair serves as a surrogate for hydrophobic drugs with limited aqueous solubility, but different results could be obtained with different cargo. FRET is used to measure micelle stability due to the sensitivity of the separation distance between donor sensitized emission of the acceptor molecule. FRET efficiency is related to the distance (FRET efficiency ~ r⁻⁶) between the donor and acceptor, thus giving insight into the location of the two dyes in relation to each other [48]. Due to this property, FRET can be used to monitor the environment of the micelle cargo and make quantitative assessements in regard to the integrity of micelle cargo and disassembly, but the relationship would not be expected to be linear when the packing of the donor and acceptor change even slightly. Subtle changes to the proximity of the dyes alter FRET efficiency, whereby a 10% increase in proximity dramatically increases (approximately 88% increase) the FRET efficiency and a 10% decrease in proximity diminishes FRET efficiency nearly in half (44% decrease). A combination of proximity changes and complete separation of the donor and acceptor can yield difficult to interpret spectral data. For this reason, complimentary methods, specifically ultracentrifugal separation in this case, are needed to understand the stability and structure of micelles in complex media.

By utilizing both FRET efficiency and ultracentrifugal separation, it becomes clear that some micelles, notably mPEG_5.4k-LA_28.5k, appear to sediment into higher density fractions following incubation with the biologic fluids, most notably in ascites, pleural, HS, and CSF. The micelle fraction appears to shift but stay intact. This may be due to association of biomolecules with the micelles causing compaction or other changes to the micelle structure, similar to results observed with solid nanoparticles [49, 50]. This remains a question to be further explored. However, both mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles appeared to dissociate with new fluorescent bands appearing in less dense fractions following incubation with biologic fluids. This change cannot be gleaned from fluorescence spectrometry alone and reveals interactions with macromolecules and extraction of micelle cargo to new less dense fractions. These changes in localization of the dyes yields confounding interpretations due to colocalization, and maintained FRET efficiency, of micelle cargo after disruption or extraction.

Our findings agree with this observation, where FRET intensity for mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles significantly decreased within the first hour upon incubation with FBS and HS. Increasing the biologic content of the biologic fluid by as little as 5% caused significant disruption of micelles [46]. This effect is described to be related to the hydrophobic block molecular weight, where higher molecular weight micelles are more stable [51]. Our results also agree with this conclusion, where mPEG_5.4k-LA_28.5k micelles exhibited a less significant decrease in FRET intensity over time along with persistence of the discrete micelle FRET bands within sucrose gradients. Shifts in the micelle density suggest that the micelles may be changing structure or even associating with macromolecules based on the micelle bands shifting into more dense gradient layers. Further research is needed to identify and characterize this phenomenon.

Association of biologic macromolecules would be expected as solid nanoparticles rapidly accumulate complement, immunoglobulin, and lipoproteins [52, 53]. Demonstrating protein interactions with micelles has been indirect due to the disruption of micelles upon mixing with proteins or during analysis [17, 36, 54]. Previous studies have utilized the hydrolysis of the micelle cargo to infer micelle stability [54] as well as the loss of sensitized FRET emissions [17]. Because of this, the research has focused primarily on disruption of the micelles by tracking their encapsulated cargo. Micelles composed of mPEG-LA exhibit stability upon exposure to γ-globulins but are disrupted in the presence of α- and β- globulins [17] whereas novel P(LA-co-TMCC)-g-PEG micelles showed an opposite relationship to these proteins [36]. Despite destabilizing the micelles, major protein constituents of the blood do not universally destabilize micelles, a phenomena that appears to be material specific [17]. Our results support this as mPEG_2.2k-CL_3.1k micelles showed dramatically higher stability in most cases compared to mPEG_2.2k-LA_2.7k, despite having a higher CMC and therefore a theoretically lower stability. This phenomenon could potentially be explained by hydrolysis of polylactide copolymers opposed to the polycaprolactone alternative. However, PLA hydrolysis at physiologic temperatures has been shown to occur over a timescale of weeks even in highly acidic conditions [55–57]. Therefore, PLA hydrolytic degradation due to pH and temperature was not considered to be a significant factor especially upon comparison between mPEG_5.4k-LA_28.5k and mPEG_2k-LA_2.7k micelles.

In the case of HS and FBS, representative of systemic administration, mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles appear to favor dissassembly as the primary mechanism of instability. To the contrary, the mPEG_5.4k-LA_28.5k micelles appear to sediment as greater density, stable bands. This is in stark contrast to mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles, where both HS and FBS caused the micelle band to decrease and disappear, suggesting dissolution. Interestingly, new less dense fluorescent bands appeared in these instances suggesting cargo extraction, uptake of both dyes into lipophilic compartments within the fluid, or association with free proteins. Taken together, micelles with smaller hydrophobic core polymer blocks underwent dissolution while micelles with longer hydrophobic core polymer blocks became more dense, which may be due to compaction of the micelles.

The disassembly of micelles was not only material dependent, but also dependent upon fluid identity. Our results show significant disruption of mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles in FBS, SF, and HS. Significant disruption was also observed in mPEG_5.4k-LA_28.5k micelles in serum and to a lesser extent in synovial fluid. For mPEG_5.4k-LA_28.5k micelles, the destabilization appeared to be independent of incubation time while mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles dissociated to a greater extent over time. Synovial fluid was found to be a highly destabilizing environment similar to serum, perhaps due to a high abundance of several classes of proteins also found in plasma, such as albumin, immunoglobulins, and apolipoproteins [58, 59]. While a lack of stability within both HS and FBS highlights the significant barrier of systemic administration, stability of micelles in other fluids suggests they are excellent candidates for alternative routes of delivery.

The mPEG_2.2k-CL_3.1k micelles were stable in pleural fluid compared to mPEG_2k-LA_2.7k micelles. Pleural fluid effusion is classified as exudate (more than 30 g/L protein) or transudate (less than 30 g/L protein) [60]. Exudate pleural fluid is considered inflammatory [61] resulting from increased vessel permeability and the accumulation of serum infiltrate within the pleural cavity [60]. This would suggest exudate pleural fluid composition to be similar to serum. The pleural fluid used in this case had high protein content and appeared to be exudate. Due to this, it was expected to have similar effect on micelle stability to that witnessed in HS and FBS for all micelles. This was not the case, suggesting that although exudate pleural fluid is derived directly from blood there is dramatically different composition and this difference may explain limited micelle stability of micelles in blood.

Despite increased protein, and other molecules from serum infiltrate, pleural fluid exhibited only mildly destabilizing conditions similar to ascites fluid, which had significantly lower protein and lipid content. Based on this, the overall protein concentration does not offer a predictive factor toward explaining dissolution of micelles with similar HLB and size. Micelles should exhibit higher stability after administration in ascites or pleural fluid without having to overcome the destabilizing effects of intravenous administration. Micelles could take advantage of ascites or pleural fluid buildup witnessed across various disease states, such as cancer and heart failure, remaining intact and increasing residence time in these tissues. This could be advantageous through reducing the dosing burden on the patient as they would be hypothesized to remain intact longer within the fluid compartment.

Unexpectedly, mPEG_2k-LA_2.7kmicelles appeared to to recover FRET signal by fluorescent spectrometry readings between 3 h and 12 h in AF (Figure 3C). Analysis of the gradient ultracentrifugation revealed the presence of separate, less dense bands increasing with time of incubation and fluid percentage (Figure 8). The micelles were not stable in these fluids, but instead had released their cargo. The FRET dye pair colocalized within new fractions, potentially lipoproteins or vesicles. This highlights the need for complementary methods when analyzing FRET-based stability assays within complex media and the need to utilize a second stability method even when it appears there is stability in a given fluid as the recovery of FRET over time could be misinterpreted by FRET spectrometry alone.

Similar to mPEG_2k-LA_2.7k micelles in AF, both mPEG_2.2k-CL_3.1k and mPEG_5.4k-LA_28.5k micelles exhibited increased FRET signal over time in CSF while mPEG_2k-LA_2.7k showed neglibile decrease in FRET compared to PBS (Figure 3). Sucrose density gradient ultracentrifugation visualization of micelles confirmed no visible appearance of separate dye bands nor qualitative decrease in micelle band intensity, providing direct evidence that CSF negligibly affected micelle stability and lack of cargo extraction. This increase in FRET could potentially indicate compaction of the micelle and decreased distance between the two dyes within the core, with significant increase in FRET efficiency with small changes to the proximity of the two dyes [48]. This idea will be further pursued by our group in the future, but the fact that the dye is not observed in other density bands leaves few other interpretations to this contradictory observation. If true, it would suggest that the mPEG_2.2k-CL_3.1k micelles have a significant change in structure (proximity of the two dyes) while the mPEG_5.4k-LA_28.5k micelles undergo only minor structural change. Regardless, all micelles examined in CSF exhibited high stability, as suggested by our prior work in this area [35].

Together, these results suggest that the identity of the fluid has a substantial effect on stability. Fuids with similar protein concentrations (FBS, HS, synovial, and pleural) had different effects on mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k stability while those with drastically different concentrations (ascites and pleural) showed similar destabilizing effects. Simply put, protein content alone was not predictive of the stability of the micelles in the fluids. Gradient ultracentrifugation results suggest that specific interactions between biologic components in the biologic fluids are driven by the specific micelles. Dissolution was favored for lactide and caprolactone containing smaller hydrophobic blocks, while an increase in density was favored for a lactide copolymer with a much larger hydrophobic block. This result contradicted expected results based on the CMC of these copolymer micelles as mPEG_2.2k-CL_3.1k micelles exhibited approximately a 5-fold larger CMC than mPEG_2k-LA_2.7k micelles (Supplementary table 2), but retained FRET dyes to a greater extent at equivalent concentrations across all fluids when separated via density and retained higher FRET ratios over time at 200x their respective CMCs according to spectrometry (Figures 2 and 3).

Appearance of separate FRET bands above the initial micelle band across FBS, HS, and synovial fluids suggests lipophilic components also disrupt micelles and solubilize their cargo. Sucrose gradient fractions at 20% sucrose to 35% sucrose correspond to densities of 1.0810 g/mL to 1.1513 g/mL, respectively. All lipoprotein classes, with the exception of HDL, have been shown to colocalize to densities up to 1.063 g/mL less dense than the 20% fraction [62]. HDL and extracellular vesicles exhibit a broader range of densities and can be separated among various fractions from 1.06 g/mL to 1.20 g/mL corresponding to sucrose fractions between 15% and 45% sucrose [63]. The colocalization of both DiO and DiI above the micelle fraction is hypothesized to be partitioning of the micelle cargo into less dense lipid vesicles or lipoproteins, much in the way dye transfer occurs to cell membranes [18].

Analysis of lipid content in pleural fluid effusion shows increased fatty acid and phospholipid production and metabolism in malignant states [64] and lipoprotein content can be modulated by the inflammatory environment of biologic fluids [65]. Cis-unsaturated lipids are common components of phosphotidylcholines within cell membranes and vesicles, accounting for >50% of all lipid species within eukaryotic membranes [66]. Therefore, the exudate pleural fluid was hypothesized to contain relatively high unsaturated lipid content, and this was supported in our chemical analysis of the fluid when compared to other fluids such as FBS, HS, and AF (Figure 1B). Fluids with high unsaturated lipid content led to the appearance of discrete fluorescent bands above the micelle FRET fractions in the cases of mPEG_2k-LA_2.7k micelles across all fluids except pleural fluid while mPEG_2.2k-CL_3.1k micelles exhibited less dramatic colocalization across the highest content fluids (SF, HS, FBS, AF) within these bands. CSF exhibited the lowest unsaturated lipid content and corresponded to no discrete fluorescent bands appearing outside the expected micelle band. Generally, the lipid content of the fluid did correlate with the fluid’s abilty to destabilize the micelles and extract their cargo. However, pleural fluid did not demonstrate a distinct new fluorescent band at lower density, suggesting that the unsaturated lipid content of the fluid alone was not a predictive factor of micelle stability. Destabilization of block copolymer micelles may be exacerbated by interactions with lipids [67, 68], but little work has been performed to characterize these interactions. There is a clear need for further characterization of the interaction between polymeric micelles and model macromolecules, such as extracellular vesicles and lipoproteins, based not only on micelle characteristics but also material physiochemical properties.

Regardless of the explanation of the stability in some biologic fluids and instability in other biologic fluids, our results identify pleural fluid, CSF, and ascites fluid as fluids in which micelles are stable even when they possess short hydrophobic chains. These results suggest, in agreement with others [17, 36] that blood is a biologic fluid in which micelles are highly unstable even at high polymer concentrations, in our case 200 times the CMC. Micelles administered locally, specifically CSF for all micelles tested and pleural or ascites fluid for mPEG_2.2k-CL_3.1k micelles, may yield much better results in terms of micelle stability, allowing for prolonged residence of micelles as well as delivery of higher concentrations of drug to the intended target. Synovial fluid destabilized all micelles, suggesting intra-articular delivery would be expected to induce a quick release of the micelle’s cargo. Ascites fluid showed drastically different effects between mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles, with ascites fluid greatly destabilizing mPEG_2k-LA_2.7k micelles while mPEG_2.2k-CL_3.1k was relatively more stable. This difference in stability could not be explained by CMC of the material alone nor protein and lipid concentration of the fluid, indicating other factors being involved.

Although these results are expected to be predictive of in vivo stability, we focused on acellular fluids which neglect the interactions with cells. We hope to expand this method to cellular environments; ultracentrifugal separation has been widely applied to separation of biologic components from cellular systems [30, 33, 34]. As discussed earlier, the differences in stability could be due to specific protein or lipid interactions with the micelles or the solubilized polymer chains. Further complicating this, the hydrophobicity of the API may also affect the intact micelle’s stability [7–9]. Therefore, care should be taken when extrapolating these results to other drug cargos as more hydrophilic cargo may exhibit altered miscibility, heat of mixing, and less favorable thermodynamic forces for micelle aggregation. As we elucidate the interactions, we will strive to determine if specific interactions are driving the instability in specific micelles. We will continue to examine the destabilizing effects that biologic fluids have on micelles, and we hope to further elucidate the mechanisms by which micelles are destabilized at high concentrations of polymer by biologic macromolecules. These results do not attempt to describe the physiologic forces these fluids undergo nor the rapid dilution of the micelles post administration [45]. Stability of micelle drug delivery systems is a complex problem and future work should examine the contributions of these factors in the context of the respective fluids.

Conclusion

Serum induces rapid dissolution of micelles and continual release of micelle cargo, whereas cerebrospinal fluid offers much greater stability for the micelles examined. Other fluids exhibited different stability depending on the diblock copolymer hydrophobic molecular weight and chemical identity. Overall, mPEG_5.4k-LA_28.5k micelles exhibited greater stability within biologic fluids compared to mPEG_2.2k-CL_3.1k and mPEG_2k-LA_2.7k micelles. The observed instability is at a concentration significantly above the critical micelle concentration, and thus dilution is not the primary factor. Both protein and lipid content of the fluids correlated with micelle instable in the fluid, i.e. higher protein and unsaturated lipid content generally correlated with lower stability. However, significant stability variations were observed even when the total protein and lipid levels were similar suggesting that specific interactions of the polymers with the biologic fluid components may be at play. These results provide insight into physiological properties that should be examined for micelle clinical development. Our results also highlight how CMC was not predictive of micelle stability in complex biologic fluids. Specifically, for two micelle system with similar HLB, zeta potential, and size, mPEG_2k-LA_2.7k micelles, with an approximately 5-fold lower CMC, were less stable than the mPEG_2.2k-CL_3.1k micelles. The mPEG_2.2k-CL_3.1k micelles were revealed to be more stable in ascites, pleural, and CSF whereas mPEG_2k-LA_2.7k micelles were stable in CSF, suggesting their suitability for local administration to these fluids. Finally, the use of FRET in combination with gradient centrifugation can allow for a better understanding of the changes taking place in micelles following incubation in biologic media and reveals the colocalization of FRET dyes in lipophilic compartments may confound FRET spectrometry conclusions.

• Human serum is one of the most destabilizing biologic fluids for polymer micelles

• Polymer micelles appear stable in human cerebrospinal fluid

• Biologic fluid total protein & lipid do not correlate with micelle instability

• Typical micelle parameters do not correlate with destabilization in biologic fluids

• Gradient ultracentrifugation complements other micelle characterization methods