Study on Doxorubicin Loading on Differently Functionalized Iron Oxide Nanoparticles: Implications for Controlled Drug-Delivery Application

Nanoplatforms applied for the loading of anticancer drugs is a cutting-edge approach for drug delivery to tumors and reduction of toxic effects on healthy cells. In this study, we describe the synthesis and compare the sorption properties of four types of potential doxorubicin-carriers, in which iron oxide nanoparticles (IONs) are functionalized with cationic (polyethylenimine, PEI), anionic (polystyrenesulfonate, PSS), and nonionic (dextran) polymers, as well as with porous carbon. The IONs are thoroughly characterized by X-ray diffraction, IR spectroscopy, high resolution TEM (HRTEM), SEM, magnetic susceptibility, and the zeta-potential measurements in the pH range of 3–10. The degree of doxorubicin loading at pH 7.4, as well as the degree of desorption at pH 5.0, distinctive to cancerous tumor environment, are measured. Particles modified with PEI were shown to exhibit the highest loading capacity, while the greatest release at pH 5 (up to 30%) occurs from the surface of magnetite decorated with PSS. Such a slow release of the drug would imply a prolonged tumor-inhibiting action on the affected tissue or organ. Assessment of the toxicity (using Neuro2A cell line) for PEI- and PSS-modified IONs showed no negative effect. In conclusion, the preliminary evaluation of the effects of IONs coated with PSS and PEI on the rate of blood clotting was carried out. The results obtained can be taken into account when developing new drug delivery platforms.


Introduction
Doxorubicin (DOX) and other antibiotics from the anthracycline series attract the attention of the scientific community due to the simultaneous combination of positive and negative effects on human body [1][2][3][4]. Furthermore, DOX plays a key role in the treatment of many neoplastic diseases, in particular cancer, even though its prolonged action causes cardiomyopathy and congestive heart failure [5,6]. A strong desire to reduce cardioand nephrotoxicity of DOX led to the idea of binding the drug to various nanomaterials, e.g., engineered nanoparticles [7]. Various nano-sized structures based on synthetic and natural polymers, solid lipids and lipid nanoemulsions (liposomes), silver, gold, calcium carbonate, metal-organic frameworks, carbon, silicon, or magnetic materials have been To reach the affected tissue, the IONs must pass through the capillaries and avoid phagocytosis, thus, their size should be less than 100 nm [83,84]. In addition, IONs must have a high magnetization which should disappear after the removal of the magnetic field in order not to cause the formation of aggregates [83,85]. Accordingly, the average size of IONs less than 100 nm or in the range of 100-150 nm is considered the best for loading and efficient drug delivery [7,86]. When choosing the proper size of IONs, it is also necessary to take into account the methods of their excretion from the body. It has been reported that particles up to 20 nm in diameter are removed from the blood through the kidneys, while particles > 100 nm, as a rule, are absorbed by the liver, spleen, or bone marrow [84].
A detailed analysis of available literature allows for several general conclusions. First, commonly used are the IONs of widely varied sizes, from small 6-10 nm, to medium (tens of nm), and quite large, over 100 nm and up to 300 nm. This is due to the fact that a priori it is unclear which particular size would provide the best targeting function [84,87,88]. For instance, for IONs ranging from 60 to 310 nm it was found that smaller nanoparticles demonstrate better cellular internalization, deeper penetration into multicellular spheroids, and provide higher efficiency of photothermal ablation in vitro. At the same time, large IONs are better accumulated in tumors, causing more effective inhibition of their growth. In their turn, 120 nm particles are best suited for magnetic resonance imaging and in vitro photoacoustic tomography. It is noted that apart from the size, the particle shape and the chemical composition of the modified surface may be important [89].
Second, in most studies the authors focus on block copolymers of complex structure (see Table S1). However, such techniques are of little use in the biomedical practice, since multi-stage synthesis is difficult and time-consuming. Furthermore, sorption and desorption processes on such modified particles sometimes last several days and have low efficiency. Third, only one type of ION surface modifier is usually subject to investigation. This shortcoming retards the choice of the most effective carrier for DOX. Likewise, little attention has been paid on the influence of the size and shape of the ION core.
The aim of this work was to (i) obtain IONs of 80-150 nm in size, (ii) modify their surface with cationic and anionic polyelectrolytes, nonionic polymer dextran, and porous carbon, and (iii) reveal the optimal core-shell system for the sorption of DOX, with the objective to develop and implement an effective nanocarrier for controlled drug delivery. The modified IONs were characterized in detail by various methods (XRD, FTIR, measuring the zeta-potential, scanning and transmission microscopy, etc.) to explain the sorption behavior.  [90,91], being in good agreement with the standard XRD data card of Fe 3 O 4 crystals (JCPDS No. . The synthesis product, in addition to magnetite, contains fractions of hematite Fe 2 O 3 and goethite FeO(OH), at 24 and 6%, respectively. High-temperature treatment (boiling) in propylene glycol reduced the proportion of hematite to 11%, while the goethite phase disappeared completely.

SEM and TEM
The results of TEM and SEM measurements of particle size are presented here and in the Supplementary Materials Figures S2 and S3. The synthesized materials are a polydisperse ensemble of crystalline particles of cubic magnetite. Particle size (d) for unmodified magnetite Fe 3 O 4 , determined from TEM and SEM images, is in the range of 80-150 nm. The calculated values of d were limited to a sample of 500 particles. The presence of fine (d < 80) and larger (d > 150) fractions does not exceed 10% by volume. Thus, the average particle size (d av ) is 115 ± 35 nm. The TEM micrographs also show variable in size aggregates with a size of 300-400 nm.
Using HRTEM, the size of the magnetite particles after surface modification (PSS, PEI, and Carb) was determined to estimate the change in the size of the shell thickness after the sorption of DOX (see Section 3.6 for more detail). Based on TEM images, it was found that the shell for PSS-and PEI-modified IONs is uniform (Figure 1c,e). The thickness of MNPs is 11.6 ± 0.6 and 11.2 ± 2.6 nm for Fe 3 O 4 @PSS and Fe 3 O 4 @PEI, respectively. The thickness of the porous carbon layer in Fe 3 O 4 @Carb (Figure 1g,h) is substantially wider, 26.0 ± 6.7 nm. The TEM micrographs witness that the IONs coated with PSS and PEI are in an aggregated state while forming the chain structures ( Figure S3). The reason may be the interaction between the polymer shells of MNPs as well as intrinsic limitations of TEM and SEM, operating in a high vacuum where particles tend to stick together.

Infrared Spectroscopy
The IR spectra of the original magnetite and magnetite coated with porous carbon and polymers are shown in Figure 2. The summary of peak assignment, given in Table 1, confirms the successful surface modification. For unmodified magnetite, the particle sizes before ( Figure 1a) and after DOX sorption (Figure 1b) are comparable. The presence of the drug was not detected on HRTEMphotographs, and the change in particle size is not significant, amounting less than 5% of the average size. The results for the modified materials also indicate the absence of noticeable changes in the particle morphology after the sorption of DOX: the shell thickness of Fe 3 O 4 @PSS@DOX was 11.0 ± 0.7 nm (Figure 1d), and for Fe 3 O 4 @Carb@DOX 26.1 ± 7.5 nm (Figure 1i). Changes for Fe 3 O 4 @PEI are more noticeable. After the sorption of the drug, the average shell thickness increased by 2.9 nm (the thickness of the Fe 3 O 4 @PEI@DOX shell was 14.1 ± 1.9, see Figure 1f), likely indicating the presence of DOX on the surface.

Infrared Spectroscopy
The IR spectra of the original magnetite and magnetite coated with porous carbon and polymers are shown in Figure 2. The summary of peak assignment, given in Table 1   Bending vibrations of C-H  Bending vibrations of C-H The IR spectra of DOX itself and magnetite nanoparticles with adsorbed DOX are presented in Figure 3. In all cases the drug's spectrum is subject to drastic changes, indicating that its sorption does take place. The strongest spectral changes are observed upon sorption on the pure magnetite nanoparticles and Fe 3 O 4 coated with porous carbon. Characteristic absorption peaks and their assignments are summarized in Table 2. The IR spectra of DOX itself and magnetite nanoparticles with adsorbed DOX are presented in Figure 3. In all cases the drug's spectrum is subject to drastic changes, indicating that its sorption does take place. The strongest spectral changes are observed upon sorption on the pure magnetite nanoparticles and Fe3O4 coated with porous carbon. Characteristic absorption peaks and their assignments are summarized in Table 2.    Stretching vibrations C=C ring bend [107] The change in the IR spectra after the sorption of DOX indicates the presence of the drug on the surface of the sorbents.

Magnetic Characterization
The magnetic properties were measured using vibrating-sample magnetometry that assumes that the saturation magnetization (Ms) is related to the sample mass [109]. By a slightest change in the Ms and coercive force (Hc), it is possible to detect not only changes in the internal structure of the magnetic material and pertinent chemical processes (particle growth, oxidation, etc.), but also changes in the mass of the sample with high accuracy [70,110].
As can be seen from Figure S4 and Table 3, implementation of the modifiers slightly reduces the magnetization compared to the parent magnetite. Specifically, functionalization with polymers, regardless of their nature, results in a magnetization decrease of almost 10%. This may indicate small differences in the thickness of the modifier layer on the magnetite surface, shielding the magnetic core and hence reducing its magnetic moment [69]. For Fe 3 O 4 @Carb, the magnetization is reduced by 20%. Importantly, for all types of nanoparticles under scrutiny, superparamagnetic behavior is observed with a characteristic absence of residual magnetization and an insignificant H C of 5.0 mT. This phenomenon presents a crucial factor for using magnetic drug carriers in the biomedical field. As already mentioned above, one of IONs' positive features is that after the removal of the external magnetic field, there is no residual magnetization, which causes the aggregation of magnetic particles. After the sorption of DOX, a slight change in the saturation magnetization was observed (Table 3) and additionally to IR-spectra and HRTEM images, this finding indicates the presence of the drug on the adsorbent and an increase in the thickness of the nonmagnetic layer of the particle. The change in Ms was no more than 4%. The greatest change is observed in PEI-and carbon-modified IONs. Results of determining the zeta-potential of functionalized IONs are presented in Figure 4. In all cases, surface modification significantly changes the ζ of the initial magnetite. For the polyelectrolyte-modified particles, the ζ exceeds the accepted stability threshold ( l. Sci. 2023, 24, x FOR PEER REVIEW 10 of 28 due to the high . The Fe3O4@PSS particles have the best stability in suspension at pH 7.4 due to the complete dissociation of modifier's sulfonic groups. Unusual increased stability in the colloidal solution of IONs coated with dextran at pH 7.4, despite the lowest values, can be explained by the steric factor [115]. On the other hand, the Fe3O4@Сarb suspension displays marginal stability (see Table 4), as an effect of large particle size and low zetapotential. Changes in the rate of colloidal stability after sorption of DOX were not observed. The value of the sedimentation rate of IONs loaded with DOX did not change significantly and was within experimental error (see Table 4, values in parentheses).
> |30| mV [111]) at most of the examined pH values so that one could expect their improved stability behavior in suspensions. Nanoparticles coated with porous carbon have a negative ζ over the entire pH range, but only at pH ≥ 10 does its value (−34 mV) correspond to the mentioned stability condition. The most plausible reason of this observation is the presence (and dissociation in an alkaline medium) of carboxyl and hydroxyl groups on the surface of porous carbon. Coating with dextran reduces the surface charge by virtue of the particle enveloping with a polymer shell, built-up via the mechanism of hydrogen bond formation [112]. At the same time, in the region of pH 5.5, an isoelectric point was observed.

Assessment of Colloidal Stability
An important characteristic for dispersion of nanoparticles is stability over time. The stability of magnetite suspensions might be low because of rather large particle sizes. Here, to assess the sedimentation stability, the method of sedimentation in a gravitational field presented in the Supplementary Materials (Procedures) was used [113]. To evaluate the effect of modification chemistry on colloidal stability, the time during which the particle sedimentation takes place was measured ( Figure S5). The data obtained, as well as the results of measuring the hydrodynamic diameter dav (Z-average size) of samples, are collected in Table 4. Table 4. Colloidal stability, zeta-potential, and hydrodynamic size of aqueous suspensions with concentration 50 mg·mL −1 at 25 °C (n = 3, p = 0.95).

Sample
Average Settling Time (min) ζ-Potential (mV) Z-Average Size (nm) a рН 5.0 a Measured immediately after evaluation of colloidal stability. b Given in parentheses is the average sedimentation time after the sorption of DOX.
The DLS measurements give only an approximate size of the MNPs because they are very sensitive to temperature, solvent viscosity, and the presence of "shell" around particles [114]. For TEM, the dried samples are to be used, since the method works under ultra-high vacuum conditions. For this reason, the average particle size obtained by DLS is usually higher [114]. Diagrams of the distribution of particle size by intensity are presented in Figure S6.
The data of Table 4 show that in all cases the stability of MNPs is not high and it is limited to 11 min. On the other hand, the tested nanomaterials demonstrate a different sedimentation rate, which indicates the important role of the surface modifier. The sedimentation time, zeta-potential, and hydrodynamic size of Fe3O4@PEI remain

Assessment of Colloidal Stability
An important characteristic for dispersion of nanoparticles is stability over time. The stability of magnetite suspensions might be low because of rather large particle sizes. Here, to assess the sedimentation stability, the method of sedimentation in a gravitational field presented in the Supplementary Materials (Procedures) was used [113]. To evaluate the effect of modification chemistry on colloidal stability, the time during which the particle sedimentation takes place was measured ( Figure S5). The data obtained, as well as the results of measuring the hydrodynamic diameter d av (Z-average size) of samples, are collected in Table 4. Table 4. Colloidal stability, zeta-potential, and hydrodynamic size of aqueous suspensions with concentration 50 mg·mL −1 at 25 • C (n = 3, p = 0.95).

Sample
Average a Measured immediately after evaluation of colloidal stability. b Given in parentheses is the average sedimentation time after the sorption of DOX.
The DLS measurements give only an approximate size of the MNPs because they are very sensitive to temperature, solvent viscosity, and the presence of "shell" around particles [114]. For TEM, the dried samples are to be used, since the method works under ultra-high vacuum conditions. For this reason, the average particle size obtained by DLS is usually higher [114]. Diagrams of the distribution of particle size by intensity are presented in Figure S6.
The data of Table 4 show that in all cases the stability of MNPs is not high and it is limited to 11 min. On the other hand, the tested nanomaterials demonstrate a different sedimentation rate, which indicates the important role of the surface modifier. The sedimentation time, zeta-potential, and hydrodynamic size of Fe 3 O 4 @PEI remain practically unchanged at both pH values and their sedimentation proceeds quite slowly due to the high J. Mol. Sci. 2023, 24, x FOR PEER REVIEW 10 of 28 due to the high . The Fe3O4@PSS particles have the best stability in suspension at pH 7.4 due to the complete dissociation of modifier's sulfonic groups. Unusual increased stability in the colloidal solution of IONs coated with dextran at pH 7.4, despite the lowest values, can be explained by the steric factor [115]. On the other hand, the Fe3O4@Сarb suspension displays marginal stability (see Table 4), as an effect of large particle size and low zetapotential. Changes in the rate of colloidal stability after sorption of DOX were not observed. The value of the sedimentation rate of IONs loaded with DOX did not change significantly and was within experimental error (see Table 4, values in parentheses). Thus, based on the results of Sections 2.1.3 and 2.1.5 on the study of the sorption of various modifiers on the surface of MNPs, as well as the literature data on the state of bare magnetite nanoparticles and polymers in an aqueous solution at pH 7.4, we can assume that this sorption is caused by the complex various types of interactions. Since, at this pH, the amino groups of PEI are partially protonated, and the OH groups on the magnetite surface are partially dissociated, its binding to the polyelectrolyte is due to both electrostatic interaction and the participation of the H-bond [91,116]. In the case of PSS, the negatively charged sulfonic groups of the polymer interact electrostatically with surface iron cations and simultaneously form H-bonds with the surface OH groups of the magnetite. The binding of dextran and porous carbon to magnetite is also caused by electrostatic forces, van der Waals interactions, and H-bonding [99][100][101][102]117]. Figure 5 shows typical sorption isotherms and dependences of the degree of sorption on the mass of the sorbent. The highest degree of sorption (65%) is achieved when using the Fe3O4@PEI nanoparticles, the outcome being in accord with their highest zeta-potential (+53 mV). It can be assumed that under these conditions, sorption takes place through the formation of hydrogen bonds between non-protonated amino groups of PEI and heteroatoms of the phenolic, alcohol, methoxy, and amino groups of DOX. According to Suh et al. [118], at pH 7.4 about 80% of amino groups of PEI, which can participate in the formation of H-bonds with DOX, are not protonated. In turn, at this pH about half of the DOX amino groups are also not protonated [119][120][121]. Our assumption is consistent with the findings of a study by Coluccini et al. [122], who by applying the 1 H NMR spectroscopy as well as NOESY and NOE NMR analysis in D2O found out that the chemical shifts of the protons of the proton-donor groups and aromatic rings of DOX, PEI are all subject to alterations upon the addition of PEI.

Sorption of Doxorubicin
A higher degree of drug loading onto Fe3O4@Carb (compared to unmodified IONs or Fe3O4@Dex) is presumably associated with the shell porous structure. Increased negative potential favors the electrostatic interaction of IONs with the drug cation formed due to the partial protonation of the DOX amino group [121]. Such interpretation is consistent with an even a higher degree of sorption (32%) by Fe3O4@PSS particles, whose negative zeta-potential at pH 7.4 is higher than that at pH 5.0. An additional factor for increasing the degree of sorption by Fe3O4@PSS may be the hydrophobic interaction of modifier's benzene rings with the hydrophobic parts of the drug molecule. From the data of Figure  5b, it follows that a 95% sorption of DOX (0.5 mg) into nanoparticulate form requires from 2 mg Fe3O4@PEI and 15 mg Fe3O4@PSS to 25 mg Fe3O4@Carb and more than 30 mg Fe3O4@Dex (all sorbents being considerably more efficient than Fe3O4).  Table  potential. Changes in the rate of c observed. The value of the sedimenta significantly and was within experim Thus, based on the results of Sec various modifiers on the surface of M magnetite nanoparticles and polyme that this sorption is caused by the com the amino groups of PEI are partially surface are partially dissociated, it electrostatic interaction and the parti the negatively charged sulfonic gro surface iron cations and simultaneous magnetite. The binding of dextran a electrostatic forces, van der Waals int 2.3. Sorption of Doxorubicin Figure 5 shows typical sorption i on the mass of the sorbent. The highe the Fe3O4@PEI nanoparticles, the outc (+53 mV). It can be assumed that und formation of hydrogen bonds betw heteroatoms of the phenolic, alcohol, Suh et al. [118], at pH 7.4 about 80% o formation of H-bonds with DOX, are DOX amino groups are also not proto the findings of a study by Colucc spectroscopy as well as NOESY an chemical shifts of the protons of the p are all subject to alterations upon the A higher degree of drug loading Fe3O4@Dex) is presumably associated potential favors the electrostatic inter the partial protonation of the DOX a with an even a higher degree of sorp zeta-potential at pH 7.4 is higher tha the degree of sorption by Fe3O4@PSS benzene rings with the hydrophobic 5b, it follows that a 95% sorption of D 2 mg Fe3O4@PEI and 15 mg Fe3O4@ Fe3O4@Dex (all sorbents being consid values, can be explained by the steric factor [115]. On the other hand, the Fe 3 O 4 @Carb suspension displays marginal stability (see Table 4), as an effect of large particle size and low zetapotential. Changes in the rate of colloidal stability after sorption of DOX were not observed. The value of the sedimentation rate of IONs loaded with DOX did not change significantly and was within experimental error (see Table 4, values in parentheses).
Thus, based on the results of Sections 2.1.3 and 2.1.5 on the study of the sorption of various modifiers on the surface of MNPs, as well as the literature data on the state of bare magnetite nanoparticles and polymers in an aqueous solution at pH 7.4, we can assume that this sorption is caused by the complex various types of interactions. Since, at this pH, the amino groups of PEI are partially protonated, and the OH groups on the magnetite surface are partially dissociated, its binding to the polyelectrolyte is due to both electrostatic interaction and the participation of the H-bond [91,116]. In the case of PSS, the negatively charged sulfonic groups of the polymer interact electrostatically with surface iron cations and simultaneously form H-bonds with the surface OH groups of the magnetite. The binding of dextran and porous carbon to magnetite is also caused by electrostatic forces, van der Waals interactions, and H-bonding [99][100][101][102]117]. Figure 5 shows typical sorption isotherms and dependences of the degree of sorption on the mass of the sorbent. The highest degree of sorption (65%) is achieved when using the Fe 3 O 4 @PEI nanoparticles, the outcome being in accord with their highest zeta-potential (+53 mV). It can be assumed that under these conditions, sorption takes place through the formation of hydrogen bonds between non-protonated amino groups of PEI and heteroatoms of the phenolic, alcohol, methoxy, and amino groups of DOX. According to Suh et al. [118], at pH 7.4 about 80% of amino groups of PEI, which can participate in the formation of H-bonds with DOX, are not protonated. In turn, at this pH about half of the DOX amino groups are also not protonated [119][120][121]. Our assumption is consistent with the findings of a study by Coluccini et al. [122], who by applying the 1 H NMR spectroscopy as well as NOESY and NOE NMR analysis in D 2 O found out that the chemical shifts of the protons of the proton-donor groups and aromatic rings of DOX, PEI are all subject to alterations upon the addition of PEI.

Sorption of Doxorubicin
A higher degree of drug loading onto Fe 3 O 4 @Carb (compared to unmodified IONs or Fe 3 O 4 @Dex) is presumably associated with the shell porous structure. Increased negative potential favors the electrostatic interaction of IONs with the drug cation formed due to the partial protonation of the DOX amino group [121]. Such interpretation is consistent with an even a higher degree of sorption (32%) by Fe 3 O 4 @PSS particles, whose negative zeta-potential at pH 7.4 is higher than that at pH 5.0. An additional factor for increasing the degree of sorption by Fe 3 O 4 @PSS may be the hydrophobic interaction of modifier's benzene rings with the hydrophobic parts of the drug molecule. From the data of Figure 5b, it follows that a 95% sorption of DOX (0.5 mg) into nanoparticulate form requires from 2 mg To gain a deeper understanding of the characteristics of the synthesized sorbents, we compared them with other reported nanosized sorbents used for loading DOX (see Table  S2). The PEI-and PSS-functionalized IONs of interest have a higher sorption capacity (691 and 325 mg·g -1 , respectively). Such a difference can be explained by the fact that in the case of PEI, the interaction with DOX involves all oxygen-containing groups of DOX and amino groups of PEI, whereas PSS can interact electrostatically only with the protonated amino group of DOX. This explains the larger amount of the PSS-sorbent required for the sorption of DOX.
Another benefit of the modified IONs is much shorter time required for making quantitative drug loading true. However, the most principal advantage comprises the potential of magnetic site-specific targeting, when the drug is delivered to the deceased tissue or organ by means of external magnetic field.

Adsorption Kinetics
Comparison of the efficiency of DOX sorption on all sorbents of our interest follows from Figure 6, with all the fitting kinetic parameters listed in Table 5. The initial adsorption stage is rapid as the DOX molecules incline to be bound onto an external, highly developed surface of the nanoparticles. The correlation coefficients (R 2 ) presented in Table 5 witness that the pseudo-second-order kinetic model describes the process of DOX adsorption more precisely. This confirms that adsorption depends on the amount of the drug and the active sites on the surface of the sorbent. It is interesting to note that the calculated values of qe, determined by the pseudo-second-order model, are larger than the experimental ones (Table 5). This may indicate the need for employing longer sorption times. To gain a deeper understanding of the characteristics of the synthesized sorbents, we compared them with other reported nanosized sorbents used for loading DOX (see Table S2). The PEI-and PSS-functionalized IONs of interest have a higher sorption capacity (691 and 325 mg·g -1 , respectively). Such a difference can be explained by the fact that in the case of PEI, the interaction with DOX involves all oxygen-containing groups of DOX and amino groups of PEI, whereas PSS can interact electrostatically only with the protonated amino group of DOX. This explains the larger amount of the PSS-sorbent required for the sorption of DOX.
Another benefit of the modified IONs is much shorter time required for making quantitative drug loading true. However, the most principal advantage comprises the potential of magnetic site-specific targeting, when the drug is delivered to the deceased tissue or organ by means of external magnetic field.

Adsorption Kinetics
Comparison of the efficiency of DOX sorption on all sorbents of our interest follows from Figure 6, with all the fitting kinetic parameters listed in Table 5. The initial adsorption stage is rapid as the DOX molecules incline to be bound onto an external, highly developed surface of the nanoparticles. The correlation coefficients (R 2 ) presented in Table 5 witness that the pseudo-second-order kinetic model describes the process of DOX adsorption more precisely. This confirms that adsorption depends on the amount of the drug and the active sites on the surface of the sorbent. It is interesting to note that the calculated values of q e , determined by the pseudo-second-order model, are larger than the experimental ones (Table 5). This may indicate the need for employing longer sorption times.
Thus, immobilization of DOX on nanoparticles may take place due to the electrostatic and van der Waals interactions, hydrogen bonds, and π-π stacking of the anthracycline DOX fragment with aromatic fragments of the coating material [122]. Thus, it is difficult to predict what factor(s) governs the efficiency and degree of immobilization.  Thus, immobilization of DOX on nanoparticles may take place due to the electrostatic and van der Waals interactions, hydrogen bonds, and π-π stacking of the anthracycline DOX fragment with aromatic fragments of the coating material [122]. Thus, it is difficult to predict what factor(s) governs the efficiency and degree of immobilization.

Desorption of DOX
An important step in developing the ION-based drug delivery system was the verification of drug release. To study the applicability of modified IONs as carriers of anticancer drugs, the behavior of DOX release was modeled at physiological (pH 7.4) and cancer cell pH (pH 5.0) The resulting kinetic profiles are depicted in Figures 7 and S7, the latter portrays the first 60 min of desorption.

Desorption of DOX
An important step in developing the ION-based drug delivery system was the verification of drug release. To study the applicability of modified IONs as carriers of anticancer drugs, the behavior of DOX release was modeled at physiological (pH 7.4) and cancer cell pH (pH 5.0) The resulting kinetic profiles are depicted in Figures 7 and S7, the latter portrays the first 60 min of desorption.
As noted in many studies, desorption of DOX at pH 7.4 proceeds 2-4 times slower than in a slightly acidic environment [10,[123][124][125]. The same behavior was observed in our experiments. As shown in Figure 7a, all samples showed no significant drug release (<7%) at pH 7.4. This collective behavior of sorbents is associated with a strong electrostatic interaction between sorbent and sorbate. Additionally, this may be due to the poor solubility of the drug itself with an increasing the pH [126]. At the same time, abrupt initial desorption ( Figure S7), which is observed in all samples during the first 15-30 min, is associated with the weak physical adsorption of some drug molecules on the surface of MNPs.
On the other hand, as the pH decreases, as shown in Figure 7b, drug desorption increases. In all cases, at pH 5.0, the drug is only partially desorbed, with a highest release of about 30% recorded for the PSS-modified IONs. It is important to mention that within the first 30 min, i.e., the timeframe simulating in vivo application of a magnetic nanoformulation, DOX release reaches up to 15%. This would generate drug active concentration of 0.14 mmol·L −1 and 2-3 times higher that could be achieved with other nanosorbents [127][128][129]. As noted in many studies, desorption of DOX at pH 7.4 proceeds 2-4 times slower than in a slightly acidic environment [10,[123][124][125]. The same behavior was observed in our experiments. As shown in Figure 7a, all samples showed no significant drug release (<7%) at pH 7.4. This collective behavior of sorbents is associated with a strong electrostatic interaction between sorbent and sorbate. Additionally, this may be due to the poor solubility of the drug itself with an increasing the pH [126]. At the same time, abrupt initial desorption ( Figure S7), which is observed in all samples during the first 15-30 min, is associated with the weak physical adsorption of some drug molecules on the surface of MNPs.
On the other hand, as the pH decreases, as shown in Figure 7b, drug desorption increases. In all cases, at pH 5.0, the drug is only partially desorbed, with a highest release of about 30% recorded for the PSS-modified IONs. It is important to mention that within the first 30 min, i.e., the timeframe simulating in vivo application of a magnetic nanoformulation, DOX release reaches up to 15%. This would generate drug active concentration of 0.14 mmol·L −1 and 2-3 times higher that could be achieved with other nanosorbents [127][128][129].
The IONs coated with Dex and Carb demonstrate relatively easy desorption in an acidic environment (but not to the full extent). In the case of Dex, two phenomena can play a role here. On the one hand, the zeta-potential (in absolute units) decreases with decreasing the pH and passes through the isobestic point at the region of pH 5.5, hence implying a weakening of electrostatic interactions and increasing the release of the drug. Similar results were observed by Liu et al. [69] for the DOX desorption from the surface of modified carboxymethylchitosan. On the other hand, a decrease in pH reduces the solubility of Dex-modified IONs (see Section 2.2) and increases their aggregation [115], which can hinder the drug release over time.
In the case of the Сarb coating, we assume the presence of two types of localization of the drug on the surface. In the first case, DOX is attached to the surface due to electrostatic forces, and a change in the zeta-potential led to a decrease in forces and desorption of the drug. In the second case, the drug enters the cavities of porous carbon layer, making desorption extremely difficult to achieve. It is possible that additional conditions are required for such desorption mechanism such as the magnetic-fieldinduced drug release [130]. The IONs coated with Dex and Carb demonstrate relatively easy desorption in an acidic environment (but not to the full extent). In the case of Dex, two phenomena can play a role here. On the one hand, the zeta-potential (in absolute units) decreases with decreasing the pH and passes through the isobestic point at the region of pH 5.5, hence implying a weakening of electrostatic interactions and increasing the release of the drug. Similar results were observed by Liu et al. [69] for the DOX desorption from the surface of modified carboxymethylchitosan. On the other hand, a decrease in pH reduces the solubility of Dex-modified IONs (see Section 2.2) and increases their aggregation [115], which can hinder the drug release over time.
In the case of the Carb coating, we assume the presence of two types of localization of the drug on the surface. In the first case, DOX is attached to the surface due to electrostatic forces, and a change in the zeta-potential led to a decrease in forces and desorption of the drug. In the second case, the drug enters the cavities of porous carbon layer, making desorption extremely difficult to achieve. It is possible that additional conditions are required for such desorption mechanism such as the magnetic-field-induced drug release [130].
In our opinion, a slow (but sufficient) release of the drug can be considered as a positive factor, ensuring its long and uniform cell-killing action. In therapeutic practice, this would allow for preventing repeated exposure, drug overdosage, and possibly reducing its negative side effects. From this viewpoint, the Fe 3 O 4 @PEI nanoparticles may have the greatest medicinal potential, with low release efficiency compensated by high sorption capacity.

Cell Viability Analysis
The cytotoxicity of DOX and IONs is well known. Therefore, our relevant interest was directed to the modifiers PSS and PEI. Whereas PSS in the free form is well studied, for instance in the treatment of hyperkalemia [131], where it showed no measurable cytotoxicity [132,133], PEI proved to be a highly toxic agent capable of initiating apoptosis and cell necrosis [134][135][136]. Moreover, as noted by Kafil et al., branched PEI is more harmful than a linear analog [134]. Furthermore, it was shown that an increase in the molecular weight of PEI inhibits the cell proliferation [137].
The widely used MTT assay has a number of limitations and is only capable of assessing cytotoxicity based on the metabolic activity of the cell according to the "livingnon-living" principle [138]. At the same time, the mechanism by which this or that agent affects the cells remains beyond the scope of the analysis. It is also impossible to transfer the properties of individual components to the entire system, which forced us to consider modifiers not as separate substances, but in combination with the core entity. The results acquired revealed that the presence of the Fe 3 O 4 @PEI and Fe 3 O 4 @PSS NPs in the cell medium (for both materials) did not statistically reduce cell survival. Figure 8 shows the percentage of living cell depending on the concentration of particles added to the cell medium. It is clear that all of the tested particles and concentrations did not affect cells viability remaining no less than 75-80%.
harmful than a linear analog [134]. Furthermore, it was shown that an increase in the molecular weight of PEI inhibits the cell proliferation [137].
The widely used MTT assay has a number of limitations and is only capable of assessing cytotoxicity based on the metabolic activity of the cell according to the "livingnon-living" principle [138]. At the same time, the mechanism by which this or that agent affects the cells remains beyond the scope of the analysis. It is also impossible to transfer the properties of individual components to the entire system, which forced us to consider modifiers not as separate substances, but in combination with the core entity. The results acquired revealed that the presence of the Fe3O4@PEI and Fe3O4@PSS NPs in the cell medium (for both materials) did not statistically reduce cell survival. Figure 8 shows the percentage of living cell depending on the concentration of particles added to the cell medium. It is clear that all of the tested particles and concentrations did not affect cells viability remaining no less than 75-80%.  The most intriguing result is the impact of small doses of IONs (5 µg·L −1 ) and PEImodified IONs (5 and 1 µg·L −1 ), as statistical analysis by non-parametric one-way ANOVA indicated their difference with the control group. Probably, this could be related to the better availability of particles for cells at small concentrations compared to applying higher concentrations, at which the particles can agglomerate and become unavailable to cells.

Real-Time Platelet Dynamics
Biocompatibility testing involves a systematic approach with many different assays. As a preliminary trial, we studied the effect of IONs on the rate of thrombus formation. Fe 3 O 4 nanoparticles, as previously found, have high biocompatibility with blood and are widely used in vitro and in vivo assays [25,139]. The presence of magnetite nanoparticles does not cause degradation of blood cells (leukocytes, thrombocytes, and erythrocytes) or noticeable side effects [140]. However, in our case, MNPs have a developed surface, imposing various interactions with blood cells. The rate of thrombus formation on the surface of glass coated with collagen was assessed during the first 10 min for three different MNPs: Fe 3 O 4 @PSS, Fe 3 O 4 @PEI, and Fe 3 O 4 . Blood without the addition of nanoparticles was used as a control.
For the control sample, the rate of thrombus formation, as well as the size of the formed thrombi, is comparable to each other (Figure 9). At the same time, the blood flow through the capillary was laminar and the speed constant. Abnormal aggregative (or other) activity was not observed. Similar results were obtained for unmodified and PSS-modified nanoparticles (see Figure 9). By the end of the analysis, the thrombi reached comparable sizes, similar to the results of the control experiment.
surface of glass coated with collagen was assessed during the first 10 min for three different MNPs: Fe3O4@PSS, Fe3O4@PEI, and Fe3O4. Blood without the addition of nanoparticles was used as a control.
For the control sample, the rate of thrombus formation, as well as the size of the formed thrombi, is comparable to each other (Figure9). At the same time, the blood flow through the capillary was laminar and the speed constant. Abnormal aggregative (or other) activity was not observed. Similar results were obtained for unmodified and PSSmodified nanoparticles (see Figure 9). By the end of the analysis, the thrombi reached comparable sizes, similar to the results of the control experiment. Figure 9. Micrographs of blood clots in the chamber in various shooting modes. DIC = differential interference contrast (optical microscope); DiOC6 = cell-permeant lipophilic dye that is selective for Figure 9. Micrographs of blood clots in the chamber in various shooting modes. DIC = differential interference contrast (optical microscope); DiOC 6 = cell-permeant lipophilic dye that is selective for the mitochondria of live cells (fluorescence microscopy); Merge = summarized image (optical and fluorescent microscopy).
Conversely, Fe 3 O 4 @PEI caused increased thrombus formation. The appearance of large aggregates of platelets and immune cells was already registered in the blood at the entrance to the flow chamber with collagen, which may indicate the effect of this sample type of particles on platelets. The morphology of the formed thrombi was visually different from other samples, and their size was much larger than in other cases.

Synthesis of IONs
Cubic magnetite (Fe 3 O 4 ) was obtained by precipitation followed by deposit aging in accordance with slightly modified published approach [141] (see Figure 10 for details). First, an alkaline solution was prepared by dissolving 3.86 g of NaOH in 400 mL of deionized water under nitrogen atmosphere. Then, a solution of iron(II) sulfate was made up by dissolving 10.0 g of FeSO 4 ·7H 2 O in 100 mL of deionized water under nitrogen atmosphere. This solution was added to the NaOH solution heated to 80 • C and constantly stirred at 2000 rpm. Next, a current of air (1000 mL·min −1 ) was passed through the mixed solution for 2 h, as a result the colloidal solution changed color from light blue to black. The resulting IONs were isolated from suspension by centrifugation (6000 rpm, 10 min), washed several times with deionized water and ethanol, and dispersed in propylene glycol under the action of ultrasound for 1 h. After heating at 180 • C for 3 h in a Teflon autoclave, the particles were cooled, washed with water and ethanol, and isolated by centrifugation (same conditions as above). The IONs were applied at a concentration of 100 µg·mL −1 in all suspensions under investigation, unless stated otherwise.

Synthesis of IONs
Cubic magnetite (Fe3O4) was obtained by precipitation followed by deposit aging in accordance with slightly modified published approach [141] (see Figure 10 for details). First, an alkaline solution was prepared by dissolving 3.86 g of NaOH in 400 mL of deionized water under nitrogen atmosphere. Then, a solution of iron(II) sulfate was made up by dissolving 10.0 g of FeSO4·7H2O in 100 mL of deionized water under nitrogen atmosphere. This solution was added to the NaOH solution heated to 80 °C and constantly stirred at 2000 rpm. Next, a current of air (1000 mL·min −1 ) was passed through the mixed solution for 2 h, as a result the colloidal solution changed color from light blue to black. The resulting IONs were isolated from suspension by centrifugation (6000 rpm, 10 min), washed several times with deionized water and ethanol, and dispersed in propylene glycol under the action of ultrasound for 1 h. After heating at 180 °C for 3 h in a Teflon autoclave, the particles were cooled, washed with water and ethanol, and isolated by centrifugation (same conditions as above). The IONs were applied at a concentration of 100 µg·mL −1 in all suspensions under investigation, unless stated otherwise.

Coating with Polyelectrolytes and Dextran
An aqueous suspension of IONs was mixed with a respective modifier solution (100 µg·mL −1 ) at a ratio of 1:10, sonicated for 10 min, and stirred for 24 h at 50 rpm. The modified particles were washed several times with water and ethanol and isolated by centrifugation at 4000 rpm ( Figure 11).

Coating with Polyelectrolytes and Dextran
An aqueous suspension of IONs was mixed with a respective modifier solution (100 µg·mL −1 ) at a ratio of 1:10, sonicated for 10 min, and stirred for 24 h at 50 rpm. The modified particles were washed several times with water and ethanol and isolated by centrifugation at 4000 rpm ( Figure 11).

Coating with Porous Carbon
Nanoparticles were coated with porous carbon (Carb) by hydrothermal carbonization in aqueous medium [117]. A total of 10 mL of an aqueous particle suspension (10 mg·mL −1 ) was added to 20 mL of water containing 200 mg of glucose, at a Figure 11. A scheme of modification of IONs with polyelectrolytes (PE) and dextran.

Coating with Porous Carbon
Nanoparticles were coated with porous carbon (Carb) by hydrothermal carbonization in aqueous medium [117]. A total of 10 mL of an aqueous particle suspension (10 mg·mL −1 ) was added to 20 mL of water containing 200 mg of glucose, at a ratio to saccharide as 1:2 by weight. The mixture was diluted with deionized water to 50 mL, stirred for 2 h and then sonicated for 20 min. The resulting suspension was quantitatively transferred to a Teflon autoclave, kept for 12 h at 180 • C, cooled, and washed several times with water and ethanol. Finally, the modified particles were separated by centrifugation at 4000 rpm ( Figure 12).

Characterization of Nanoparticles
The size and shape of nanoparticles were measured using a scanning electron microscope with a resolution of 1.2 nm at 30 keV, the VEGA 3 Tescan (Brno, Czech Republic) and the transmission electron microscope Tecnai-G2 F20 (FEI Company, Hillsboro, OR, USA) with a spatial resolution of up to 0.2 nm. Hydrodynamic diameter (z-average size) and surface charge (zeta-potential, ζ) values of sample suspensions were recorded by dynamic light scattering (DLS) analysis using a Nano-ZS Zetasizer, model ZEN3600 (Malvern Instruments Ltd., Malvern, UK) at an angle of 173°. X-ray diffraction measurements were carried out on a Bruker D2 PHASER diffractometer (Bruker, Billerica, MA, USA) using Cu Kα radiation (λ = 0.154 nm) at 40 kV and 30 mA in the range of 2θ values (from 10° to 80°). Absorption and IR spectra were recorded with a Shimadzu UV-2550 spectrophotometer (Shimadzu, Kyoto, Japan) and a Bruker Vector 22 FTIR spectrometer (Bruker, Ettlingen, Germany), respectively. Magnetization curves were monitored using a Lake Shore 7407 magnetometer (Lake Shore Cryotronics Inc., Westville, IN, USA). The solutions were mixed on Bio RS-24 analog controlled rotator (BioSan, Riga, Latvia). Used for the magnetic separation was a permanent Nd-Fe-B magnet with (VN)max = 40 MGOe (Guangzhou, China). For dispersion and functionalization of IONs, a Grad 57-35 serial ultrasonic bath with a generator power of 165 W (Grad-Technology, Moscow, Russia) was used. For hydrothermal carbonization, a 50 mL Teflon autoclave with a Toption (Xi'an, China) stainless steel body was employed. Sedimentation analysis and sample centrifugation were performed, respectively, on an Vibra HT-224RCE analytical balance Shinko Denshi, Co. (Tokyo, Japan) and using an EBA 200 centrifuge (Hettich, Kirchlengern, Germany).

Sorption of DOX
An analysis of the literature showed that the optimum conditions for the sorption of DOX correspond to neutral phosphate-buffered saline (PBS ; Table S2). Polypropylene tubes were used for sorption, as previously recommended [42]. To assess the effect of sorption time, 5 mL of standard ION suspension and 5 mL of aqueous solution of doxorubicin (100 µg·mL −1 ) were placed in each of six test tubes and the mixture was diluted to 30 mL with phosphate buffer solution at pH 7.4 (10 mmol·L −1 ). The tubes were

Characterization of Nanoparticles
The size and shape of nanoparticles were measured using a scanning electron microscope with a resolution of 1.2 nm at 30 keV, the VEGA 3 Tescan (Brno, Czech Republic) and the transmission electron microscope Tecnai-G2 F20 (FEI Company, Hillsboro, OR, USA) with a spatial resolution of up to 0.2 nm. Hydrodynamic diameter (z-average size) and surface charge (zeta-potential, ζ) values of sample suspensions were recorded by dynamic light scattering (DLS) analysis using a Nano-ZS Zetasizer, model ZEN3600 (Malvern Instruments Ltd., Malvern, UK) at an angle of 173 • . X-ray diffraction measurements were carried out on a Bruker D2 PHASER diffractometer (Bruker, Billerica, MA, USA) using Cu Kα radiation (λ = 0.154 nm) at 40 kV and 30 mA in the range of 2θ values (from 10 • to 80 • ). Absorption and IR spectra were recorded with a Shimadzu UV-2550 spectrophotometer (Shimadzu, Kyoto, Japan) and a Bruker Vector 22 FTIR spectrometer (Bruker, Ettlingen, Germany), respectively. Magnetization curves were monitored using a Lake Shore 7407 magnetometer (Lake Shore Cryotronics Inc., Westville, IN, USA). The solutions were mixed on Bio RS-24 analog controlled rotator (BioSan, Riga, Latvia). Used for the magnetic separation was a permanent Nd-Fe-B magnet with (VN) max = 40 MGOe (Guangzhou, China). For dispersion and functionalization of IONs, a Grad 57-35 serial ultrasonic bath with a generator power of 165 W (Grad-Technology, Moscow, Russia) was used. For hydrothermal carbonization, a 50 mL Teflon autoclave with a Toption (Xi'an, China) stainless steel body was employed. Sedimentation analysis and sample centrifugation were performed, respectively, on an Vibra HT-224RCE analytical balance Shinko Denshi, Co. (Tokyo, Japan) and using an EBA 200 centrifuge (Hettich, Kirchlengern, Germany).

Sorption of DOX
An analysis of the literature showed that the optimum conditions for the sorption of DOX correspond to neutral phosphate-buffered saline (PBS ; Table S2). Polypropylene tubes were used for sorption, as previously recommended [42]. To assess the effect of sorption time, 5 mL of standard ION suspension and 5 mL of aqueous solution of doxorubicin (100 µg·mL −1 ) were placed in each of six test tubes and the mixture was diluted to 30 mL with phosphate buffer solution at pH 7.4 (10 mmol·L −1 ). The tubes were placed in a rotary mixer and stirred at 20 rpm. At certain time intervals (2,5,10,15,30, and 45 min), one test tube was removed and after the precipitation of IONs by applying an external magnetic field, the supernatant was separated to determine the unbound drug spectrophotometrically [142].
To assess the effect of the nanosorbent mass, 8 solutions were used, containing the same amount of doxorubicin (500 µg) and increasing amounts of IONs, from 0.5 to 50 mg (final volume 10 mL). The tubes were placed in a rotary mixer and stirred for 45 min at 20 rpm. The IONs were then magnetically separated and the concentration of DOX in the supernatant determined ( Figure 13). an external magnetic field, the supernatant was separated to determine the unbound drug spectrophotometrically [142].
To assess the effect of the nanosorbent mass, 8 solutions were used, containing the same amount of doxorubicin (500 µg) and increasing amounts of IONs, from 0.5 to 50 mg (final volume 10 mL). The tubes were placed in a rotary mixer and stirred for 45 min at 20 rpm. The IONs were then magnetically separated and the concentration of DOX in the supernatant determined ( Figure 13). The degree of sorption was calculated as where С0 is the initial concentration of the drug and С is its residual concentration after sorption. To compare the efficiency of drug sorption by different sorbents, we used the sorption capacity calculated by the following equation: where q is the difference between the initial (m0) and residual (m) masses of DOX in solution and mS is the mass of the sorbent. Each experimental series was repeated at least five times. We repeated the measurements three times and built graphs of the averaged values.

Adsorption Kinetics
To evaluate the adsorption kinetics, we used the pseudo-first-order and pseudosecond-order models [109] specified by the following equations: and Figure 13. Exploring doxorubicin sorption on modified IONs. DOX = doxorubicin.
The degree of sorption was calculated as where C 0 is the initial concentration of the drug and C is its residual concentration after sorption. To compare the efficiency of drug sorption by different sorbents, we used the sorption capacity calculated by the following equation: where q is the difference between the initial (m 0 ) and residual (m) masses of DOX in solution and m S is the mass of the sorbent. Each experimental series was repeated at least five times. We repeated the measurements three times and built graphs of the averaged values.

Adsorption Kinetics
To evaluate the adsorption kinetics, we used the pseudo-first-order and pseudosecond-order models [109] specified by the following equations: and where q t is the amount of adsorbed drug at time t per 1 g of sorbent (mg·g -1 ), q e is the equilibrium (maximum) amount of adsorbed drug per 1 g of sorbent (mg·g -1 ), k 1 (min -1 ), and k 2 (g·mg -1 ·min -1 ) are the pseudo-first and pseudo-second order constants, respectively.

Desorption of Doxorubicin
In the release studies, 5.0 mg of differently functionalized IONs with adsorbed drug were placed in 50 mL of phosphate buffer (pH 5.0 and 7.4) and mechanically stirred for 24 h at 20 rpm. At the pre-determined time points (5,10,15,30,60,120,360, and 1440 min), a 5 mL aliquot was removed, the supernatant was separated from the sorbent using a magnet and the doxorubicin concentration was determined by the spectrophotometric method. DOX releases were calculated using the following equation: where m t is the amount of released DOX at time t and m ∞ is the total amount of DOX loaded onto the sorbent. Each experimental series was repeated at least three times and the averaged values were used to build graphs.

Cell Viability Analysis
Cell viability analysis was carried out according to the method described earlier [143], with slight changes.

Cell Viability
The effects of the nanoparticles on N2A cells were determined by the standard Alamar-Blue test. Cells were seeded into 96-well cell culture plates (Eppendorf, Hamburg, Germany) at a cell density of 10 4 /well in the culture medium and incubated at 37 • C under 5% CO 2 during 24 h. The particle solutions at concentrations 1, 5, 10, 50, and 100 µg·L −1 were tested. After 24 h, 10 µL of the fluorescent dye AlamarBlue (10,000 U·mL −1 , Thermo Fisher Scientific, Waltham, MA, USA) was added to each well, and the fluorescence (540/590 nm) intensity was measured with a spectrophotometer (Infinite F200 PRO, Männedorf, Switzerland).

Real-Time Platelet Dynamics Ex Vivo Observed by Confocal Microscopy
The rate of thrombus formation in the presence of NPs was assessed according to a slightly modified method described in [144].

Human Blood Collection
Human blood collection investigations were performed in accordance with the Declaration of Helsinki under a protocol approved by the NMRC PHOI Ethical Committee Informed written consent was obtained from all donors. Blood was collected in sterile tubes containing Hirudin anticoagulant (100 U·mL −1 ; Merck, Darmstadt, Germany).

Preparation of Flow Chamber System and Human Blood for Perfusion
Glass coverslips (24 × 24 mm; Heinz Herenz Medizinalbedarf GmbH, Hamburg, Germany) were cleaned with Plasma Cleaner (Harrick Plasma Inc, Ithaca, NY, USA). In order to immobilize collagen fibers on the surface of cleaned glass coverslips, we diluted stock solution of native Chronolog fibrillar type I collagen reagent (1 mg·mL −1 , Chrono-Log, Havertown, PA, USA), which already has collagen fibrils, with a 20 mM acetic acid (Reakhim) to a final concentration of collagen solution of 200 µg·mL −1 . This solution was added as a 10 µL drop to the surface of a cleaned glass coverslip and incubated for 40 min in a humid chamber at room temperature, then rinsed with water, dried, and assembled as part of the parallel platelet flow chamber described in [145]. Before perfusion, the fluorescent dye DiOC 6 (3) (3,3 -dihexyloxacarbocyanine iodide; Thermo Fisher Scientific) was added in an amount of 0.1% (final concentration 1 µM) of the blood volume and an aqueous dispersion of IONs (concentration 0.5 µg·µL −1 ) in an amount of 1% of the blood volume.

Conclusions
Thus, in this work, we functionalized the nanomagnetite surface with cationic (PEI), anionic (PSS), and nonionic (dextran, Dex) modifiers, as well as porous carbon (Carb), and compared the degree of loading of these and bare MNPs with doxorubicin. Using DLS and TEM, we found that the size of the obtained sorbents is in the range of 80-290 nm. We have shown that modification with polymers increases the stability of particles in aqueous suspension by a factor of 4-6 compared to unmodified magnetite. A comparison of the sorption capacity of MNPs with respect to DOX made it possible to establish the following series of sorption efficiency: Fe 3 O 4 @PEI (691) > Fe 3 O 4 @PSS (325) > Fe 3 O 4 @Carb (151) > Fe 3 O 4 @Dex (63) > Fe 3 O 4 (45) mg g −1 . All samples showed no significant drug release (<7%) at pH 7.4, while at pH 5.0 the release rate and extent of desorption, although increased by 2-3 times, remained relatively low. The drug release percentage varies from 30 for Fe 3 O 4 @PSS to 13-16 for Fe 3 O 4 @Dex and Fe 3 O 4 @Carb, and up to about 6% for Fe 3 O 4 @PEI after a 30 min incubation at pH 5.0. Cytotoxicity analysis showed high biocompatibility of Fe 3 O 4 , Fe 3 O 4 @PSS, and Fe 3 O 4 @PEI. The survival rate of Neuro2A cells was above 80%. A preliminary assessment of the effect of unmodified and PSS-modified IONs on the rate of thrombus formation in the blood did not reveal noticeable changes. We believe that the obtained materials and results of the study can be taken into account when developing new drug delivery platforms.

Institutional Review Board Statement:
The study was performed in accordance with the Declaration of Helsinki under a protocol approved by the FRCC CHOI Ethical Committee.
Informed Consent Statement: Written informed consent has been obtained from all donors to publish this paper.
Data Availability Statement: Not applicable.