Uptake of polymeric nanoparticles in a human induced pluripotent stem cell-based blood–brain barrier model: Impact of size, material, and protein corona

The blood–brain barrier (BBB) is a selective barrier protecting the brain from circulating components such as toxins and pathogens in the blood.^1–3 By regulating the paracellular transport with cell–cell contacts, which include tight and adherent junctions of neighboring endothelial cells,^4,5 a transport to the brain is possible only via specific transporters.^3,6 A high transendothelial electrical resistance (TEER) and a low permeability for small molecules above 500 Da and more than five proton donors^3,6–8 distinguish this tight microvascular barrier. Unfortunately, this function leads to a challenging treatment of, e.g., neurodegenerative diseases and brain tumors as potential drugs struggle to cross the barrier.⁹ 

Due to their small size, high drug load, and the decrease of a drug’s toxicity in peripheric tissue,^10,11 nanoparticles (NPs) have been seen as drug delivery systems for a successful transport of potential drugs into the brain.^12–18 Additionally, polymeric NPs made of polylactide (PLA) or poly(lactic-co-glycolic acid) (PLGA) are biodegradable¹⁹ and approved by the U.S. Food and Drug Administration (FDA) for pharmaceutical use in humans.²⁰ A successful uptake of NPs made of PLGA and PLA with transferrin, insulin, and a low-density protein receptor has been observed.²⁰ By modifying the NP surface with a surfactant like polysorbat 80 (Tween® 80) or a receptor ligand, an improvement in the NP uptake efficiency was achieved.²⁰ 

The addition of NPs in a biological environment results in the adsorption of proteins developing a protein corona on the particle surface.²¹ Having an impact on particle toxicity, cell uptake, and agglomeration,^21,22 the protein corona is influenced by physicochemical characteristics such as nanoparticle size, material, and surface charge.^23,24 Furthermore, the biological and physiological environment as well as the incubation time also contribute to its composition.^25,26 As one of the main components in blood, albumin can commonly be found as a component.²⁶ Fibrinogens, apolipoprotein E, and complement component C3 are also often bound to nanoparticles’ surfaces.^26,27 Nevertheless, the protein corona does not reflect the composition of its biological origin.²⁶ Polymeric proteins, in particular, have a large number of different apolipoproteins in their corona,²⁸ and their number is even increased when using Tween 80 as a surfactant.²⁹ 

Next to the characterization of NPs and their potential to enhance the uptake of drugs into the brain, these effects have been often studied using BBB models, which do not reflect the human in vivo condition due to missing barrier properties, particularly the corresponding tightness. Primary cells of rats, mice, and pigs have been used,^30–32 while not representing the human physiology.³³ However, cells from human origin should be used to further improve preclinical testing and to reduce evitable animal experiments at an early stage of investigations. Due to the reported limits of human cell lines,³⁴ human induced pluripotent stem cells (iPSCs) have been seen as an alternative for establishing an in vitro BBB model^35–37 for improving translational research. Because of their pluripotency, iPSCs have been successfully differentiated into human brain microvascular endothelial cells (hBMECs) expressing all relevant transporters, receptors, as well as adherent and tight junction proteins building a tight barrier with a maximum TEER of up to 4000 Ω cm² in monoculture.^35,38 So far, most studies of iPSC-based BBB models have focused on the standardization of the differentiation protocol as well as the characterization of the model.^38,39 To date, the number of studies looking at the biological characterization of NPs as drug delivery systems to the brain using iPSC-based in vitro BBB models is still limited, and observations reported by using immortalized cell lines or animal-based model systems should be verified using these advanced and now available model systems.^40,41

Having a more realistic and reproducible model of human origin, the aim of this study was to develop efficient protocols to show the potential of iPSC-based BBB models for the biological characterization of NPs and to study the influence of nanoparticle size, material, and their protein corona in correlation with the uptake of polymeric nanoparticles. In contrast to the most prominently used endothelial cell-based BBB models, an iPSC-based model with a significantly high TEER approaches in vivo conditions where a TEER of up to 8000 Ω cm² has been measured.^42,43 This allows us to investigate a possible impact on the barrier integrity resulting from overall characterized particles by an advanced and broad spectrum of physicochemical analyzing techniques. Biodegradable polymers, poly(lactic-co-glycolic acid) (PLGA) and poly(d,l-lactide) (PLLA), have been intensively studied and approved by the U.S. Food and Drug Administration for pharmaceutical use in humans.^20,44 Thus, Tween 80-stabilized NPs made from these polymers and in the sizes of 50 and 100 nm were used in the present study.

Unconditioned medium (UM) was prepared as described in Lippmann et al.³⁷ and consisted of the following: DMEM/F12 (Gibco by Life Technologies), 20% Knockout Serum Replacer (Gibco by Life Technologies), 1× MEM Nonessential Amino Acids (Gibco by Life Technologies), 1 mM Glutamax (Gibco by Life Technologies), and 0.1 mM β-mercaptoethanol (Carl Roth GmbH). After sterile filtration, the media were stored at 4 °C.

The endothelial medium (EC) comprised an endothelial basal medium (Promocell GmbH) with 1% platelet-poor plasma-derived serum (Alfa Aesar). Forty-eight hours before and after the first 24 h of subcultivation, 20 ng/ml of basic fibroblast growth factor (bFGF; Sigma-Aldrich) and 10 μM retinoic acid (RA; Sigma-Aldrich) were supplemented. After sterile filtration, the media were stored at 4 °C for up to 2 weeks.

iPSC iPS(IMR90)-4 (WiCell) were maintained on Matrigel^© (Corning) and cultivated with mTeSR^™ Plus (Stemcell Technologies). The medium was changed daily. For a cultivation preserving the pluripotency of the cells without changing the media on weekends, the cells were seeded at a ratio of 1:10 in double media volume. When a confluence of 70% was attained, the cells were passaged. The cells were washed with phosphate-buffered saline (PBS) and detached by an incubation with Versene (Gibco by Life Technologies) for 5 min at 37 °C. After centrifugation at 400 g for 5 min, the cells were collected in fresh media and seeded at a ratio of 1:3 to 1:10.

For the differentiation of iPSCs into hBMECs, an optimized protocol of Lippmann et al. and Neal et al. was used.^37,39 After reaching a confluence of 70%, the differentiation was initiated by the cultivation of the cells in UM for 7 days with a daily change of media (weekends excluded). This was followed by a cultivation without daily change of media in EC with RA and bFGF. After 48 h, the cells were collected with Accutase (Capricorn) for 30–45 min and 1 000 000 cells/cm² were seeded in fresh EC with RA and bFGF on transwell filters or cell culture plates coated with 400 μg/ml of rat collagen I (Ibidi GmbH) and 100 μg/ml of fibronectin (F. Hoffmann-La Roche AG; 200 μl for a 24-well transwell filter) overnight. bFGF and RA were removed 24 h after the passage, and the media were not changed afterward.

The TJ-associated proteins, occludin, PECAM/CD31, and claudin-5, were detected via immunocytochemical staining. Cells were washed with PBS and fixed with 4% paraformaldehyde in PBS (Alfa Aesar) for 20 min at room temperature (RT). Afterward, the cells were blocked for 1 h with 1% bovine serum albumin (BSA; Sigma-Aldrich) in PBS and permeabilized with 0.1% Triton^™ X-100 (Sigma-Aldrich). The cells were washed three times with PBS and the primary antibody diluted in 1% BSA in PBS (1:25 PECAM/CD31, 1:50 occludin, and 1:100 claudin-5) was added for 1 h at RT. After rinsing the cells three times, the secondary antibody (1:1000 in 1% BSA in PBS; Alexa Fluor 594 antirabbit for PECAM/CD31 and occludin, Alexa Fluor 594 antimouse for claudin-5) was added for 1 h at RT. The cells were washed three times with PBS and nuclei were stained with Hoechst 33342 (1:5000 in PBS) for 5 min at RT. Following a three-time wash with PBS, 1–2 drops of ProLong Glass Antifade Mountant (Thermo Fisher Scientific) for each well were added. After an overnight incubation at RT, the cells were visualized using a wide field microscope (Leica AF7000 or Olympus BX51).

For the measurement of the TEER in Ω cm², CellZscopeE (nanoAnalytics GmbH) and EVOM2 (World Precision Instruments Inc.) were used. Transwell filters for 24-well plates (Corning) were coated with 400 μg/ml of rat collagen I and 100 μg/ml of fibronectin overnight at 37 °C. After washing the transwell filters with PBS, 1 000 000 cells/cm² in 350 μl fresh EC-media were seeded on the filters. The lower compartment was filled with 1 ml of fresh EC media. Coated transwell filters without cells were used as the control. After inserting the transwell filters in the CellZscopeE, the TEER was measured automatically every hour.

For measuring the TEER with EVOM2, the chopstick electrodes were sterilized by incubation in 70% ethanol for 15 min. First, the controls without cells were measured. The TEER was calculated by multiplication of the resistance with the area of the used transwell filter and subtraction of the control. The TEER was measured every 24 h.

The permeability of the brain endothelial cell layer was studied with sodium fluorescein (Sigma-Aldrich) with a size below 4 kDa. Forty-eight hours before adding sodium fluorescein, the cells were seeded on transwell filters. After reaching a TEER of at least 1500 Ω cm², the medium in the upper compartment was changed with 350 μl EC media with a final concentration of sodium fluorescein at 10 μg/ml. Every hour, 3 × 50 μl were removed from the lower compartment and transferred to a 96-well plate, and 150 μl of fresh medium were added to the lower compartment afterward. For the standard curve, dilutions of the stock solution starting at a ratio of 1:100 and continuing at a ratio of 1:4 were prepared. The fluorescence was measured in the fluorescence reader at an excitation wavelength of λ = 485 nm and an emission wavelength of λ = 535 nm. The standard curve was used to determine the concentration of sodium fluorescein. The permeability coefficient was calculated with the following equation:

where V_A is the volume of the lower compartment (cm³), A is the area of membrane (cm²), C_D is the final concentration of sodium fluorescein (μg/ml), d_CA/d_t is the increasing concentration in the lower compartment over time [(μg/ml)/min], and P_app is the permeability coefficient (cm/min).

Nanoparticles were synthesized by a combination of miniemulsion and solvent evaporation process.^45,46 Biodegradable polymers, RESOMER® R 202 H (PLLA) and RESOMER RG 502 H (PLGA; both from Evonik), were used. All NPs have the fluorescent dye N-(2,6-diisopropylphenyl)-perylene-3,4-dicarbonacidimide (PMI; BASF) encapsulated in them and are stabilized by the nonionic surfactant Tween 80. For the preparation of 100 nm NPs, 300 mg of PLLA or PLGA and 0.22 mg of PMI were dissolved in 10 g of chloroform. The macroemulsion was prepared by adding the aqueous phase consisting of dissolved 60 mg of sodium dodecyl sulfate (SDS) in 24 g of water to the organic phase and subsequent magnetic stirring of the mixture at 1000 rpm for 60 min. Afterward, the macroemulsion was sonicated under ice cooling for 180 s at 70% amplitude in a pulse regime (30 s sonication and 10 s pause) using a Branson 450W sonifier and 1/4 in. tip. The obtained miniemulsion was transferred to a 50 ml reaction flask with a large-sized neck and left under gentle stirring overnight at 40 °C for complete evaporation of chloroform. After evaporation, SDS was exchanged with Tween 80 through three times centrifugation and redispersion of NPs in 1 wt. % Tween 80 aqueous solution (after the first two centrifugations) and in demineralized water after the last centrifugation. For the preparation of NPs with the average size of 50 nm, 150 g of PLLA or PLGA and an aqueous phase consisting of 72 mg of SDS dissolved in 24 g of water were used. The rest of the process parameters were identical.

The average size of NPs was determined by dynamic light scattering (DLS) using a Nanoflex DLS (Microtrac Europe GmbH, Germany) with a measuring angle of 180° and de-ionized water as dispersion medium.

All asymmetrical flow field-flow fractionation (AF-FFF) measurements were carried out on an AF-FFF V6 System from Consenxus GmbH. The system consisted of a constaMETRIC® 3200 pump for the main flow, a Knauer WellChrom Micro-Star K-100 pump for the injection flow, an AF-FFF channel V6 equipped with a 200 μm poly(methyl methacrylate) channel plate and a regenerated cellulose membrane with a molecular weight cutoff (MWCO) of 5 kDa, a Bronkhorst Hi-Tec mini CORI-FLOW^™ mass flow controller to control the cross flow, and a Flowbox V6 from Consenxus GmbH, containing all valves and electronics. The samples were measured by application of an injection loop with a volume of 25 μl. A Waters 486 UV absorption detector, set up for a wavelength of λ = 254 nm, as well as a Consenxus Dark V6 light scattering (LS) detector set up to measure the scattered light intensity at an angle of 90°, was used for the sample detection. The eluent used for the measurements of the samples without prior incubation in fetal calf serum (FCS) consisted of de-ionized water containing 5 mmol/l NaCl and 50 μl/l Tween 20 as a surfactant. For the measurements of samples after incubation with FCS, PBS was used as an eluent.

To be able to derive particle sizes from the fractograms, which are originally obtained from the FFF measurements, a calibration of the system is necessary.⁴⁷ Here, the calibration was conducted using five different standards with particle sizes in the range of the investigated polymer nanoparticles. As size standards consisting of PLLA or PLGA were not available, polystyrene (PS) standards were used to derive a PS-equivalent calibration. Two different methods were used for the AF-FFF measurements, because the retention time of the samples after incubation with FCS shifted toward significantly longer retention times and the sample peaks showed a strong tailing. This led to elution times outside of the measurement range of the first method [Fig. 1(a)]; therefore, a different method [Fig. 1(b)], with a different flow profile and longer measuring time, had to be applied for the measurements of the samples with prior incubation in FCS. The background for this necessity was that retention times in AF-FFF should generally be kept as short as reasonably possible. Aside from practical benefits, this is of significance, as unnecessary long retention times lead to undesirable band broadening.^48,49 Table I shows the sizes of the used polystyrene standards.

To perform the calibration, the standards were diluted to a mass concentration of 0.25 mg/l with de-ionized water prior to injection. Figure 1 shows the size distributions of the polystyrene standards for each AF-FFF method (after application of the calibration).

For the analysis of the PLLA and PLGA nanoparticle samples, these samples were diluted to a mass concentration of 1 g/l with de-ionized water. Additionally, AF-FFF analyses were performed after incubation of the particulate samples with FCS. For the AF-FFF analyses, incubation of the polymer particles with FCS was conducted as follows: The nanoparticle dispersions were mixed with de-ionized water (see Table II for the amounts), and 25 μl FCS were added. After 10 min, 25 μl of a concentrated PBS solution (3M) were added. As a result, the samples contained a final concentration of 1 g/l for the particles, 5 vol. % FCS, and the standard electrolyte contents for PBS (≈150 mM). The samples were incubated for 2.5 h at room temperature.

For the development of a protein corona on the nanoparticle surface, 0.1 m² of NP surface in 150 μl NP solution were incubated with 500 μl blood plasma (a mixture of 20 individual samples) for 1 h at 37 °C under rotation at 500 rpm. NPs incubated with PBS were used as the control. The following equation was used for the calculations:

where S_1gNP is the surface of 1gNP (m²), ρ is the density of polymer material (g/m³), and d is the diameter of NP (nm).

By filling up to 1 ml and vortexing, the reaction was stopped. Nonadsorbed proteins were removed by centrifugation at 4 °C and 20 000g for 30 min, which was repeated four times. The final pellet was redispersed in 1 μl PBS and stored for mass spectrometry analysis at −80 °C.

A matrix of the ten most frequently proteins from each sample was generated. After subtraction of the controls, the percentage of these proteins of each sample was determined.

To exclude the autofluorescence of the cells, the controls were subtracted from the measurements. Due to differences in the encapsulated amount of the fluorescent dye PMI in the nanoparticles, the fluorescence of the NP was different. To allow a comparison of the different NP samples, the fluorescence signal of the NPs was normalized to the sample with the highest fluorescence. The fluorescence intensity (FI) of the NP samples in triplets at equal concentrations was measured and the factor was calculated with the following equation:

The cells were incubated with NPs at a concentration of 150 μg/ml for 20 h at 37 °C. To analyze the NP uptake, the supernatant (media and PBS) of the following steps was retained until centrifugation. After washing the cells with PBS, the cells were detached using Accutase®. The reaction was stopped by adding fresh EC media. After centrifugation for 5 min at 400 g, the cells were collected in 200 μl Fixable Viability Stain 660 (FVS660; BD Biosciences; 1:1000 in PBS) and incubated for 15 min at RT. For the flow cytometric (BD Accuri C6) analysis of the NP uptake, the cells were excited at a wavelength of 488 nm. Channel FL1 (filter 533/30) was used to determine the fluorescence intensity of the NPs. Decayed cell compartments were excluded by setting gates with 20 000 events as one event represents a cell. Applying bd accuri c6 software for the statistical analysis, the median of the measurements of each NP in channel FL1 was used. The following equation was used to calculate the relative NP uptake in comparison with the NP with the highest FI:

Binding to amines of necrotic cells, FVS660 was used to determine the level of cell viability (an excitation wavelength of 640 nm and an emission wavelength of 660 nm in channel FL4) simultaneously. Cells without NPs were used as the control.

The toxicity of the NPs was analyzed with Cell Counting Kit-8 (CCK8) and crystal violet. First, NPs were added to the cells at different concentrations (three wells per concentration) for 20 h at 37 °C. As a control, cells without NPs, as well as cells lysed by a treatment with 0.1% Triton X-100 for 1 h (positive control), were used. Afterward, the medium was replaced with CCK-8 solution according to the manufacturers’ protocol for 1 h at 37 °C. The solution was transferred to a new 96-well plate and the absorption was measured at λ = 450 nm. Afterward, the cells were fixed with ethanol/methanol (2:1) for 1 h at RT and stored in PBS at 4 °C.

For the analysis via crystal violet, 50 μl per well of a 0.1% crystal violet solution were added to the fixed cells for 20 min at RT and 70–80 rpm. Excessive dye was removed by washing the cell-containing wells with water and drying the wells overnight. By adding 100 μl 33% acetic acid for 10 min at RT and 70–80 rpm, the dye was dissolved from the cells. A total of 90 μl of the dissolved dye were transferred to a new 96-well plate and the absorption was measured at λ = 600 nm. For the evaluation, the mean of the media was subtracted from the single measurements and the cells without NPs were set to 100%.

Only partially expressing tight junction-associated and endothelial proteins, in vitro models of the BBB based on human cell lines often do not generate a tight barrier with physiologically relevant barrier properties.⁵⁰ Unfortunately, the access to human primary cells is very limited, particularly to cells derived from healthy donors. Being available from autopsies or biopsies from diseased brain tissues, the risk of the isolated cells being contaminated with tumor cells is real and high, which do not make them a sustainable option for establishing reproducible BBB models for screening purposes.³⁷ iPSCs have been seen as an alternative, as their pluripotency allows the differentiation of the cells into hBMECs expressing endothelial as well as tight junction-associated proteins and relevant membrane transporters, resulting in a tight barrier with physiological properties.^35,37–39

Herein, the differentiation was done according to Lippmann et al.³⁷ and Hollmann et al.³⁸ and initiated by the cultivation in UM for 7 days.³⁷ A cultivation in EC with RA and bFGF is followed for 2 days, promoting the development of endothelial cells.³⁷ RA is secreted by astrocyte progenitors promoting the expression of VE-cadherin, tight junction proteins, occludin and claudin-5, as well as efflux transporters, MRP and Pgp, leading to a higher yield of cells building a tight barrier.^51,52 To ensure enough supplements for the cells, the medium was replaced daily in the established protocols.^35,37,38 Different from these previous protocols,^35,37,38 the medium was not replaced on weekends during the cultivation of iPSCs as well as the differentiation period. Nevertheless, a successful differentiation of hBMECs expressing tight junction proteins was achieved using the new protocol, resulting in a tight barrier built by a monolayer of endothelial cells.

Due to the differentiation leading to a mixture of endothelial cells and immature neuronal cells, hBMECs were further selected by subcultivation on a fibronectin-collagen matrix. So far, all protocols have used human collagen IV for the development of endothelial progenitor cells and the differentiation process,^35,37–39 because collagen IV is the main component of the basal membrane, which is secreted by brain endothelial cells in vivo.⁵³ In contrast to the already published protocols, rat collagen I, which is successfully used for the cultivation of other human brain microvascular endothelial cells (e.g., hCMEC/D3),⁵⁴ was utilized, and this also resulted in the successful differentiation into hBMECs [Figs. 2(a)–2(c)]. An obvious advantage of rat collagen I is not only its lower price, but also its general availability compared with human collagen, making the applicability of the established protocol easier and comparable over time.

The differentiation of iPSCs into hBMECs was finally verified via the positive immunocytochemical staining of the endothelial protein PECAM/CD31 as well as the tight junction-associated proteins, occludin and claudin-5 [Figs. 2(a)–2(c)]. The expression of these proteins is indicative of tight cell–cell contacts of brain endothelial cells preventing a paracellular transport of small molecules. The results in Figs. 2(a)–2(c) show a strikingly diffuse staining of intracellular structures in addition to the staining of the membranous proteins. Greene et al. demonstrated that during neuronal development, claudin-5 is mainly expressed in the cytoplasm.⁵⁵ A similar observation was also made with occludin.⁵⁶ Combined with additional isoforms, this can also lead to unspecific immunocytochemical staining.⁵⁷ Hence, the diffuse staining that occurred in the present cell differentiations can be explained by the protocol mimicking the neuronal development.

A successful differentiation is also distinguished by a high TEER and a low permeability. Thus, the TEER was measured and it increased over time up to 4000 Ω cm² [Fig. 2(e)] for up to 5 days of hBMECs cultured on transwell filters with a pore size of 0.4 μm. Using different protocols, iPS(IMR90)-4 have shown a maximum TEER of 3000 Ω cm² staying over 1500 Ω cm² for 4–10 days, as demonstrated by Lippman and others.^38,39,51 Overall, all models obtained a significantly increased TEER in contrast to porcine endothelial cells with 347 Ω cm²,⁵⁸ hCMEC/D3 with 40 Ω cm²,⁵⁹ primary endothelial cells with 340 Ω cm²,³⁷ and periphery endothelial cells with 3–30 Ω cm².³⁷ For the analysis of NP transport, a cultivation of the cells on transwell filters with a pore size of 3 μm, which reduces the attachment of the NPs on the filters, is mandatory to study a potential transport across the BBB model and the pores of the filter membranes to the lower compartment.¹² Although the TEER is reduced to 2000 Ω cm², no intrusion of the barrier integrity was observed [Fig. 2(e)].

Indicating barrier functions, the high TEER measurements correlate with low permeability measurements. After 1 h of adding sodium fluorescein, the permeability is at 6 × 10⁻⁵ cm/min [Fig. 2(d)]. After 2 h, the permeability decreases, stays constant at 4 × 10⁻⁵ cm/min, and is 100 times lower than in the controls without cells. The permeability coefficient of 4 × 10⁻⁵ cm/min significantly shows differences from the control with 2 × 10⁻³ cm/min correlating with the low permeability of hBMECs at 10⁻⁵ to 10⁻⁶ cm/min with a TEER above 1500 Ω cm² in other BBB models.^38,39,51 In comparison, the extensively used cell line hCMEC/D3 with 5.5 × 10⁻³ cm/min has a 100 times higher permeability for sodium fluorescein.⁶⁰ 

Due to the observed properties and the characteristics intensively described by others,^35,37,39 this iPSC-based BBB model enables investigations of potential impacts on the barrier properties by the uptake of drugs and drug delivery systems such as NPs. Improving such model systems may lead to a more detailed understanding of the uptake of drugs and NP–cell interactions as well as of brain delivery vehicles by NP translocation studies, enabling the transfer of the results from the lab to real applications more easily.

Due to their small size, variations in their shape, and their modifiable surface, inorganic as well as polymeric NPs are able to cross the BBB.¹² Particularly, polymeric NPs have the advantage of being biocompatible, variable in their polymer composition, and able to encapsulate a variety of drugs.¹² Thus, NPs made of the biodegradable polymers, PLLA and PLGA, were used for this study. Size characterization of the synthesized PLGA and PLLA nanoparticles was performed using DLS. Additionally, zeta potential measurements were conducted. The results of the size and zeta potential characterization are given in Table III.

When nanoparticle characterization is done in an electrolyte-poor medium or in media with isotonic salt content only (without the presence of proteins), the results obtained under such conditions are of poor biological relevance. Therefore, the AF-FFF analyses described within this publication were conducted using PBS as an eluent after incubating the samples in FCS to obtain a better comparability of the applied conditions to in vivo conditions.⁶¹ AF-FFF measurements deliver real-size distribution data due to the fact that the measurement principle is based on the fractionation of a sample. In contrast, to retrieve size distribution data from DLS measurements, assumptions and models have to be applied; most prominently, a strictly spherelike particle shape is assumed, which is, at least in most cases, not strictly true (e.g., also for agglomerating particles). Additionally, bigger particles within a size distribution are overestimated due to their disproportionally larger scattering intensity. Due to these (and other) facts, size distribution data derived from DLS measurements comprise a not negligible uncertainty.

For comparability reasons, AF-FFF analysis runs were performed using the FCS mixture as the only analyte under identical conditions as they were applied later when measuring the FCS-incubated particulate samples. Therefore, FCS was diluted to 5 vol. % with de-ionized water and analyzed by AF-FFF. With these measurements, one obtains a reference for the elution time and the amount of serum proteins as they have to be expected to also elute during a measurement with particles incubated in 5% FCS—at least under the premise that there are no additional interactions between serum components and the particles. Figure 3 shows the fractogram of FCS diluted to 5% with de-ionized water. Because most serum components are smaller than the size of the smallest size standard used for the calibration, in this case, only the fractogram is given and not a size distribution. The low signal of the LS detector confirms the expectation that light scattering is not a particularly well-suited method to detect very small particles such as proteins.

Figure 4 shows the fractogram of PLLA-50 particles after being incubated with 5% FCS in PBS for 2.5 h at room temperature. The before shown measurement with 5% serum in de-ionized water has been added to be able to compare the amount of free serum in the sample containing the PLLA-50 particles. This comparison yields that the serum concentration is slightly larger, although a slight decrease should be expected due to the formation of a protein corona around the particles.^62,63 This observation is likely caused by a coelution of the surfactant Tween 80, which is used to stabilize the nanoparticle dispersions, with the unbound proteins. Furthermore, the comparison of the measured signal intensities again shows that the UV detector is better suited to detect the presence of proteins compared with the light-scattering detector, whereas the light-scattering detector is better suited to detect the nanoparticles.

Figure 5 shows the comparison of the size distributions of the PLLA-50 particles before and after incubation with 5% FCS in PBS. As already mentioned, two different methods had to be applied when performing the underlying measurements to account for the large increase in retention time after the incubation with 5% FCS. Therefore, these measurements can only be compared after “translation” of the fractogram into size distributions as it is done by application of the calibration data. However, it was decided to plot the size distributions against relative particle sizes in this case. This decision was made for the following reasons: In AF-FFF, the calibration should optimally be conducted with the same material as the material under investigation and under exactly the same conditions (meaning, e.g., the concentration of electrolytes, surfactants, and other additives), because different materials and additives may show different interactions with the membrane in the FFF channel and with the analyte particles. This could lead to different retention times for two particles with the same size but consisting of different materials or even for two particles of the same size and material but measured in the presence of different additives.^64–66 In the given case, the measurements of the unincubated particles had to be performed in an eluent different from that for the measurements of the incubated particles. Although the effect of the application of different methods for the measurement of the particles can, in a simple case, be compensated for by the application of the calibration data, it is not possible to achieve meaningful absolute sizes due to the remaining circumstances in the given case. Since the measured signal intensities for the two measurements were vastly different, it was additionally decided to normalize the signal intensities to achieve better comparability of the size distributions.

From the normalized relative size distributions, one can derive that after incubation with 5% FCS, the particle peak is shifted toward longer retention times and is significantly broader than the reference measurement containing only PLLA-50 particles (see also Fig. 5). This is indicative of the formation of a protein corona around the particles, which alters the retention properties of the particles.⁶¹ The measurement results can be explained by the assumption that the protein corona around the particles changes the way the particles interact with the membrane, as an additional surface layer with different properties obviously alters the surface properties.²³ 

Table IV shows the relative size increase after the incubation with 5% FCS in PBS measured with AF-FFF. The data show a size increase after incubation with an FCS of around 50%, except for the PLGA-100 particles, which showed a significantly smaller size increase of only 30%. The reason for this observation is a subject of further investigations. Graphs corresponding to Figs. 3 and 4 for the remaining three materials are shown in the supplementary material.⁸⁹ 

Due to the adsorption of proteins on the particle surface, the incubation of NPs in a biological medium leads to the development of a so-called protein corona.^67,68 Influenced by physicochemical characteristics such as size, material, and surface charge,^69,70 the protein corona often has a significant impact on the toxicity, cell uptake, and agglomeration of NPs.^21,22 Therefore, the compositions of the protein corona of the four different particles were analyzed via mass spectrometry (Fig. 6). The used plasma [Fig. 6(a)], a mixture of 20 individual samples, was also analyzed and showed a composition of 50% albumin. Other dominant proteins were pregnancy zone protein (13%), alpha-1 antitrypsin (10%), and tissue factor (9%). However, the protein corona composition does not reflect the composition of the blood plasma. Although albumin constitutes 50% of the serum, its percentage in the protein corona is only between 6% and 10. These results confirm the selectivity of the protein adsorption in other studies with silica and PLGA particles,^23,28,71 which can be retraced to their physicochemical characteristics.⁷² Herein, this effect was shown the first time for PLLA and PLGA nanoparticles with diameters of 50 and 100 nm, respectively.

The composition of the protein corona is also determined by the size, surface charge, material, and stabilization of NPs.^23,24 Nevertheless, the studies here do not show significant changes in the protein corona composition of NPs with different sizes and polymers [Fig. 6(b)]. However, the use of Tween 80 for the stabilization of all NPs could be an explanation. Previous work has shown that the stabilization of NPs with Tween 80 leads to a protein corona primarily dominated by apolipoproteins,^12,73,74 being at 33%–42% in PLLA-NP and 13% in PLGA-NP in this study. Furthermore, complement components also make up a high percentage of the protein corona supporting the results of Allemann et al. and Shah and Singh,^28,75 who also analyzed polymeric NP made of PLLA and PLGA. Here, the percentage of complement components in the protein coronas of PLLA particles is at 16%–19% and at 18%–22% in PLGA particles.

The composition of the protein corona also has an influence on the NP uptake at the BBB.²⁵ Dominating the protein corona, apolipoprotein is one of the ligands of low-density lipoprotein receptors of endothelial cells allowing a receptor-mediated uptake at the BBB.⁷⁴ As a potential uptake mechanism, this ensures a successful uptake of NPs at the BBB. A modification of the NP surface with target ligands such as transferrin has been used to enhance the uptake of NPs at the BBB in other studies.⁷⁶ Unfortunately, the effect of an enhanced uptake by these target ligands can be covered due to proteins being adsorbed on the particle surface.⁷⁷ Instead, a surface modification with surfactants like polysorbat 80 (Tween 80), which was performed here, leads to a protein corona consisting of a high percentage of target ligands like apolipoprotein E.^29,78 All NPs have been internalized by hBMEC, showing the potential of all nanoparticles to be used as drug delivery systems for targeting human brain endothelial cells (Fig. 7). However, differences in the uptake efficacy, toxicity, and impact on the tight barrier of the BBB could be observed despite not having significant differences in the composition of the protein corona (Figs. 7 and 8).

In addition to the physicochemical differences of the particles, studies by Müller et al. and Yan et al. observed a denaturation of proteins of the protein corona on the particles’ surface.^79,80 The denatured proteins lose their functionality and cannot interact with potential receptors to ensure a successful uptake of the NPs, resulting in a reduced uptake.

Furthermore, the existence of a hard and a soft corona has been discussed. By the mass spectrometry analyses presented here, only the hard corona, meaning those proteins with a high binding affinity,^69,70,81 was analyzed. So far, the number of studies about the soft corona has not been many,²⁶ and, therefore, it is not known whether the soft or the hard corona is primarily responsible for a new biological identity of NPs.⁶³ In addition, the denaturation of the adsorbed proteins, which can differ depending on the size and material of the particle,⁸² might affect the biological identity of the studied NP. This could lead to differences in the NP uptake, toxicity, and their impact on barrier integrity despite having no significant differences in their protein corona compositions.

Showing no significant differences in the composition of the protein coronas of NPs in different sizes and materials, it could be assumed that the observations of those NPs in cell uptake, toxicity, and barrier integrity also do not differ. Due to the successful differentiation of iPSCs into hBMECs, these cells were used to analyze those aspects. In contrast to most other studies using TEER measurements to study the impact of NPs on barrier integrity, the TEER was measured using the CellZscopeE. Due to the fixated position of the electrodes and the automatic measurements every hour, variances in measurements based on the position of the electrodes or the time frames are excluded. Using in vitro BBB models based on animal cells and a TEER significantly below 1000 Ω cm², others studying the influence of NPs on barrier integrity did not see a significant reduction in the TEER.^83–85 In the present study, after NP incubation at a concentration of 150 μg/ml, only the uptake of PLGA-100, which had the lowest size increase after incubation in 5% FCS (Table IV), leads to a reduction by 52%–67%, differing significantly from cells not incubated with NPs [P < 0.0021, not shown; Figs. 7(a) and 7(b)]. Yet, a complete collapse of the barrier integrity cannot be observed. With a particle size of 50 nm, the TEER is reduced by 7%–10% after NP addition in comparison with cells without NPs. Nevertheless, no significant changes could be observed. Cells not treated with NPs attained a TEER between 2800 and 1500 Ω cm² similar to cells with PLLA-100.

Furthermore, the NP uptake was analyzed by flow cytometry [Figs. 7(c) and 7(d)]. Due to different encapsulation efficiencies of the fluorescent dye, the fluorescence signal was normalized to the sample with the highest fluorescence, PLLA particles (100 nm). Therefore, the relative NP uptake is shown in comparison with this sample. By comparing PLGA particles sizing between 114 and 250 nm and a negative surface charge, particles below 250 nm are taken up 90% more⁸⁶ promoting the often shown conclusion of a higher uptake of smaller sized NPs.¹² Here, the analysis of the NP uptake via fluorescence-based flow cytometry shows that the uptake of PLGA-NP with sizes of 50 and 100 nm are both significantly 1.8 times higher than PLLA-NP. Thus, a material-dependent rather than a size-dependent increase of NP uptake was perceived.

However, the analysis of the NP uptake via flow cytometry does not detect only those cells that have successfully taken up the cells within their cytoplasm but also cells that have only absorbed NPs on their surface. Hence, the NP uptake was verified via fluorescence microscopy [Figs. 7(e)–7(h)]. NP incubation for 20 h led to such a high uptake that cell structures could not be observed. Therefore, the incubation time was reduced to 4 h. Showing the intracellular uptake of all NPs, the microscopic analysis confirms the results of the flow cytometry of a successful uptake of all particles. Agglomeration has been observed, especially with NPs ranging from a size of 21–190 nm as well as at high concentrations.⁸⁷ Nevertheless, all NPs, except PLGA-NPs at a size of 100 nm, build agglomerates showing a possible correlation with the reduction in TEER and a lower size increase after incubation with FCS.

The combination of the high uptake with a significant reduction of the TEER of 100 nm-sized PLGA particles may lead to the assumption that the particles may have a toxic effect on the cells. Therefore, the toxicity of the particles was studied in detail (Fig. 8). FVS660 stains necrotic cells and is analyzed simultaneously to the NP uptake via flow cytometry. The results show that the particles are nontoxic at the given concentration [Figs. 8(a) and 8(b)]. Further toxicity analysis was done to verify the observed results. CCK-8 relies on the metabolic reduction of WST-8 to formazan. Crystal violet is a dye intercalating in DNA and proteins, measuring the cell number. Driven by the concentration gradient,^3,6–8 the uptake via transcellular diffusion is higher for NPs with a size under 50 nm.^12,74 The high NP concentration leads to a high accumulation of NPs within the cells, causing toxic effects.⁸⁸ This toxicity can be observed in a reduced metabolism of the cells, which was analyzed via CCK-8 and crystal violet. Looking at the results of the CCK-8 analysis [Figs. 8(c) and 8(d)], only PLGA-NPs with a size of 50 nm at a concentration of 1000 μg/ml significantly cause severe toxicity in the cells, whereas NPs with a size of 100 nm cause only mild toxicity. At lower concentrations, the toxicity of PLGA-NPs at a size of 50 nm is mild with 60%–70% of living cells. NPs with a size of 100 nm show no toxicity at concentrations below 1000 μg/ml. Although the percentage of living cells increases up to 40%, the results of the toxicity analysis with crystal violet [Figs. 8(e) and 8(f)] correlate with those of CCK-8. PLGA-NPs with a size of 50 nm and a concentration of 1000 μg/ml are severely toxic for the cells. With a size of 100 nm at the same concentration, the NPs are mildly toxic. No toxicity can be observed at lower concentrations for both NP sizes. It should be noted that crystal violet only stains proteins and DNA of adherent cells. Additionally, crystal violet may interact with proteins on the particles, which results in an increased staining intensity.

Nevertheless, none of the NPs show toxicity at a concentration of 150 μg/ml in all performed assays, which was used for studying the barrier integrity. Thus, the significant TEER reduction after the uptake of PLGA-NPs at a size of 100 nm cannot be retraced to their toxicity, but rather to the combination of their high uptake, their minor increase in size after incubation with FCS, and their level of agglomeration.

Filling the gap between already existing immortalized cell culture models and animal experiments, the iPSC-based BBB model shows a potential for studying the NP influence on barrier integrity. The combination of using a more suitable BBB model with a complemented characterization of NP properties increases the potential of successfully translating new NPs into future medical applications, which enable the treatment of brain diseases such as tumors. Furthermore, AF-FFF enables the investigation of size and agglomeration of NPs in biologically relevant media, which increases the value of particle characterization and supports understanding NP–cell interactions. This overview indicates the ability of Tween 80-stabilized PLGA and PLLA particles with sizes of 50 and 100 nm to be used as potential delivery systems to microvascular endothelial cells within the brain. Correspondingly, further studies are required to verify these hypotheses.

This work was financially supported by the German Federal Ministry of Education and Research BMBF (Project “nanoBBB,” Contract No. 13N13529), the German Research Foundation (DFG) via the Collaborative Research Center SFB 1066, and the Evonik Stiftung.

The data that support the findings of this study are available from the corresponding author upon reasonable request.