Antibody-Loading of Biological Nanocarrier Vesicles Derived from Red-Blood-Cell Membranes

Antibodies, disruptive potent therapeutic agents against pharmacological targets, face a barrier in crossing immune systems and cellular membranes. To overcome these, various strategies have been explored including shuttling via liposomes or biocamouflaged nanoparticles. Here, we demonstrate the feasibility of loading antibodies into exosome-mimetic nanovesicles derived from human red-blood-cell membranes, which can act as nanocarriers for intracellular delivery. Goat-antichicken antibodies are loaded into erythrocyte-derived nanovesicles, and their loading yields are characterized and compared with smaller dUTP-cargo molecules. Applying dual-color coincident fluorescence burst analyses, the loading yield of nanocarriers is rigorously profiled at the single-vesicle level, overcoming challenges due to size-heterogeneity and demonstrating a maximum antibody-loading yield of 38–41% at the optimal vesicle radius of 52 nm. The achieved average loading yields, amounting to 14% across the entire nanovesicle population, with more than two antibodies per loaded vesicle, are fully comparable to those obtained for the much smaller dUTP molecules loaded in the nanovesicles after additional exosome-spin-column purification. The results suggest a promising new avenue for therapeutic delivery of antibodies, potentially encompassing also intracellular targets and suitable for large-scale pharmacological applications, which relies on the exosome-mimetic properties, biocompatibility, and low-immunogenicity of bioengineered nanocarriers synthesized from human erythrocyte membranes.

Accordingly, significant research efforts are being devoted to devising efficient methodologies for the delivery of antibodies across immune system and cell membranes, ranging from intracellular injection to camouflaged transport techniques. 14The former relies on harsh mechanical disruption of the cell membrane through injection or electroporation, with limited loading efficiency and significant impact on cell viability, exclusively suitable for in vitro studies. 15,16 lternative approaches involve antibody camouflaging using cell-penetrating peptides, engineered nanoparticles, or liposomes to facilitate antibody transport across cellular membranes. 5, 6, 9-13, 17, 18.Among these, nanocarrier-assisted delivery, employing polymeric nanoparticles, 5,19 lipid nanovesicles, 20 and nanoparticles camouflaged with the aid of biomimetic coatings derived from cell membranes [21][22][23][24][25] stand out as promising approaches for drug delivery.Liposomes, known for biocompatibility and controlled release properties, face limitations due to protein corona formation and short-term cargo preservation effects. 26Some challenges can be mitigated by PEG-polymerization, 19 which however may trigger anti-PEG immunoglobulin production in-vivo, resulting in lowered blood circulation times and degraded immunogenicity. 27,28  innate biocompatibility and non-immunogenicity of red blood cell (RBC) membranes makes them ideal raw materials for direct use as bio-camouflaging materials in a variety of treatments and as drug carriers for intracellular delivery, in nanovesicle forms. 24,29,30 RC membrane-coated nanocarriers have been already studied for Ab delivery and proven to afford longer circulation times thanks to functional RBC-membrane proteins such as CD47.22,23,28,29,31 However, the use of such RBC-camouflaged nanocarriers requires the Ab cargo to be aggregated first into a solid form, 21,25 which may compromise its functionality and induce complications.32 Such drawbacks can be overcome by drug carriers directly synthesized from RBC-membranes in the form of nanovesicles, provided that suitable procedures become available for their loading with antibodies.33,34 Recently a novel methodology was devised for synthesizing and loading RBC-derived nanovesicles, similar to exosomes, enabling large-scale production in stable formulations with engineerable properties.This technique, initially applied for loading dUTP cargo molecules 35 , has now been extended to demonstrate the loading of RBC membrane-derived nanovesicles with larger molecular cargos, such as goat-anti-chicken IgY (H+L) secondary antibodies with significantly larger molecular weights (~145 KDa) than labeled dUTP (~1 KDa).This study systematically analyzes and quantitatively compares the results of Ab-loading with dUTP-loaded vesicles under identical processing conditions, employing spectroscopic protocols developed for single-vesicle profiling with single-molecule resolutions.33,35 The findings reveal that Ab-loading yields are maximized for slightly (∼5-10 nm) larger vesicle radii than the ones of dUTP-loading, consistent with the smaller size of the latter, yet still in the ∼50 nm radius range typical of exosome-mimetic nanocarriers. Additional cleaninf nanocarrier solutions using an exosome spin column shows comparable average loading yields of 14% for both Ab and dUTP.The inferred average number of cargo molecules loaded in each nanovesicle features also very similar values (2.25 for Ab and 2.49 for dUTP), exceeding two in both cases, despite their large size discrepancy.The results provide clear evidence for the viability of human erythrocyte-derived nanovesicles for Ab-loading and pave the way to their exploitation as a novel biomimetic system for potential antibody delivery.
Figure 1a provides a flow chart for the preparation of antibody-loaded nanovesicles, closely following previously developed procedures for dUTP cargo molecules. 36The method involves multiple ultracentrifugation steps to purify RBC ghosts from human blood, 37 and isolate exosomelike vesicles from detergent-resistant membrane (DRM) solutions at a buoyancy of 30% sucrose (1.13 g/cm3) in sucrose gradients (details in S1).The Ab-loading is performed through posthypertonic lysis of RBC vesicles, 38 inducing vesicle rupture and their subsequent revesiculation, upon which they may engulf Ab-molecules deliberately dispersed in physiologic buffer (see also S1).For this pilot study, a goat-anti-chicken IgY (H+L) antibody is chosen as the cargo molecule.
The antibody is conjugated with AlexaFluor®488 (Thermo Fisher), for green fluorescence tagging in optical characterizations using dual-color fluorescence microscopy (DCFM).The outer membranes of the nanovesicles are further stained with CellVue®Claret (Sigma-Aldrich), for farred fluorescence.The dual-color green and red tagging scheme is illustrated in Fig. 1b.As highlighted in Fig. 1a, the sample preparation process encompassed two slightly different sample typologies of nanovesicles, denoted as RBC and RBC + .The main difference between the two consists in an additional cleaning step performed at the end on the latter (RBC + ), with an exosome spin column purification procedure, 39 as detailed in S1.In all cases, loaded and tagged RBC or RBC + nanovesicles, dispersed in PBS solution, underwent systematic characterizations by means of atomic force microscopy (AFM, Fig. 1c-d) and confocal fluorescence microscopy (Fig. 1e-f), according to experimental and analytical protocols originally defined in previous publications. 35 The considerable heterogeneity of the exosome-mimetic nanovesicles under investigation necessitates characterizations at the single-vesicle level to extract critical physical parameters, such as size distribution, and facilitate thorough assessments of antibody loading.To address these challenges, we employ experimental methodologies that involve concurrent AFM and DCFM measurements (Fig. 1c-f) and dual-color coincident fluorescence burst (DC-CFB) analyses for size-resolved characterizations of both carrier nanovesicles and their cargos.Figure 2 presents key outcomes of these characterizations conducted on the overall populations of RBC and RBC + nanovesicles following their synthesis and Ab-loading procedure.][43] As illustrated by the plots of Fig. 2a and b, and further quantified by the data in Table I DCFM experiments, as outlined in S1, provide comprehensive insights into nanovesicle populations and their loading. 35The burst analysis of time traces from the red membrane dye (Fig. 1b) in the fluorescence experiments (Fig. 1c) allows for the retrieval of size-dependent statistics for the entire nanovesicle populations, as depicted in Fig. 2c-d.Fluorescence-derived estimates for size of nanovesicles tend to be slightly larger than those obtained by AFM.This discrepancy is attributed to the larger hydrodynamic size of vesicles in physiological solution, the setting for fluorescence microscopy measurements, compared to dry conditions used for AFM.The observed shift in the peak radius of fluorescence and AFM distributions is relatively small (  −    ~ 5 ).However, the difference becomes more pronounced when considering the average values of vesicle radii (  −    ~ 30 ) are likely due to the random diffusion trajectories through the detection volume and the tendency of biomimetic nanovesicles to aggregate in physiological solutions, as consistent with prior reports. 35This effect, documented in the literature, 44 is confirmed by the longer tails in the distributions obtained from fluorescence data, particularly visible for  ≫ 50  in Fig. 2c-d  A notable observation from comparing Ab and dUTP loading results is the larger vesicle size associated with antibody loading.This aligns with the substantial weight difference between Ab molecules and dUTP, with the former being over two orders of magnitude higher-molecular weight than the latter.A distinct difference is observed in the values of   and   after applying the extra cleaning procedure (RBC vs RBC + ) to dUTP-loaded samples, an effect not observed in antibody loading.In the dUTP case, both   and   show an increase of approximately 5-7 nm post-cleaning.This suggests additional size-filtering effects during the exosome spin column process, potentially favoring slightly larger vesicles and better matching the size-distribution peak (  ) of Ab-loaded samples, which is approximately 5 nm larger than in the dUTP case.The impact of cleaning (RBC vs RBC + ) is more pronounced in the case of dUTP loading compared to Ab loading, significantly affecting also the retrieved loading yields, as discussed in the next section., quantified as the ratio of the number of loaded nanovesicles (  ), determined from coincident green and red bursts, (Fig. 3a-b) and the total nanovesicle count (  ), determined from red burst analyses (Fig. 2c-d), as a function of the nanovesicle radius .Table II summarizes key figures of merit extracted by the DC-CFB analysis of the experiments, to enable quantitative comparisons between RBC and RBC + preparations, as well as loading with Ab and dUTP cargo molecules.Similar to dUTP-loaded nanovesicles, the Ab-loaded nanovesicle populations exhibit a skewed distribution in radius, with a primary peak at    < 70 nm and an extended tail toward larger sizes (>100 nm), where experimental artifacts due to vesicle agglomeration combined with random diffusion trajectories become prominent.The size distributions of Ab-loaded nanovesicles in RBC (Fig. 3a) and RBC + (Fig. 3b) samples show essentially the same values for the peak (   =52 nm) and average (   ∼65 nm) radii, indicating negligible impact of the solution cleaning step.
Comparing the histograms in Fig. 3a-b with those of Fig. 2c-d case.However, this is not observed for the dUTP case, as evident in the loading yield distributions for RBC and RBC + preparations (dashed lines in Fig. 3c and 3d) and the corresponding figures of merit in Table II.The RBC + cleaning step induces a clear modification of the peak yield, with    increasing from 42 to 48 nm and    decreasing from 67% to 52%, along with substantial effects on average values, with    increasing from 55 to 65 nm and   decreasing from 20% to 15%.These trends align with those highlighted in the vesicle populations for dUTP cargo molecules, indicating a more pronounced influence of the additional cleaning process and its associated size-filtering effect, as discussed with reference to Fig. 2   Equivalent data for the dUTP case are presented in S3 (Fig. S11), and a comparative summary of the results is provided by Table II (columns 6 and 7), listing the retrieved values of the maximum number of cargos per loaded nanovesicle (  ) and its average (  ) for all four sample typologies.interaction distance of hemoglobin molecules are unpredictable and uncontrollable, leading to varying lifetimes.To account for this effect on the fluorescence brightness of the tagging dye inside RBC and RBC + vesicles (see also S2), the area under their normalized lifetime histograms (a2 and a3 in Fig. 4c and d) is compared with that of free cargo molecules (a1 in Fig. 4c-d), allowing for the calibration of the average photon count for the entrapped cargos (see Table S3).
Conclusively, the DC-CFB assessments provide the two-dimensional profile of loaded vesicles versus radius and Ab-number (number-normalized brightness) per loaded vesicle, as illustrated in Figure 5, revealing the mostly populated sizes and the load extent for the Ab-loaded nanovesicles.In summary, we have demonstrated successful antibody-loading into synthesized nanovesicles from red-blood-cell membranes.Dual-color fluorescence burst investigations at single-vesicle level revealed a preference for nanovesicles with an average radius of around 65 nm when loaded with Ab-cargos, compared to 55 nm obtained in reference experiments performed on dUTP cargos, which aligns with the larger size and molecular weight of Ab molecules.Loading yields for Abcargos were comparable to those of dUTP-loaded vesicles, peaking at approximately 40%.The optimal vesicle radius of 52 nm and an average loading efficiency of 14% was obtained for Abloaded nanovesicles.Considering unexpected lifetime shortening of Alexa488 fluorophore in produced nanovesicles, likely due to FRET interactions with hemoglobin molecules, revealed an average loading of 2.25 antibody molecules per nanovesicle, consistent with dUTP cargo results under identical conditions.Additional cleaning procedures also provided stable results for Ab-loading and insights into the relationship between nanovesicle physical properties and cargo nature, offering opportunities to enhance production yields and customize these nanovesicles for specific therapeutic Ab-agent delivery.

ASSOCIATED CONTENT
Supporting Information.Supporting information is available free of charge at http://pubs.acs.org and includes details on the experimental methods, data analysis procedure and summary of additional reference results.were pelleted in low sucrose (1-3%) by ultracentrifugation at 100,000g and 4°C for 1h (BC, SW32Ti) to remove excess fluorescence.

d) Red fluorescence staining of DRM-nanovesicles
In Figure S1, the process for staining loaded nanovesicles with red fluorescence is highlighted within a red box.To label the vesicle membranes, the pellets were mixed with CellVue Claret membrane kitstaining-component Diluent (Sigma-Aldrich, Miniclaret-1kt), following the manufacturer's instructions.The staining reaction was stopped by 3% bovine serum albumin in PBS and immediately pelleted by top-loading on a low concentrated sucrose solution (1-3%) at 100,000g and 4°C for 1h (BC, SW32Ti).The pellets of DRM-nanovesicles preparation were resuspended in PBS and kept at 4°C in dark for physics to be conducted.Short summary, this approach is applied to investigate the efficiency of Ab loading of the prepared samples and compare them with dUTP cargo.To reach this goal, the cargo molecules were fluorescently stained by the dye Alexa488 with excitation and detection wavelengths at 485 nm and 535 nm, respectively, while nanovesicle membranes were tagged by the far red dye, CellVue Claret, with corresponding excitation and detection wavelengths of 640 nm and 720 nm, respectively.

e) Extra cleaning with exosome spin column
In the last phase of the preparation, to evaluate the impact of additional purification on the loading efficiency and cleanliness of loaded RBC nanovesicles, a portion of these loaded DRM-nanovesicles underwent a purification process utilizing the exosome spin column (ESC, ®Thermo Fisher, ESC MW 3000).
In Figure S2

S2-Data analysis
The dual-color coincident fluorescence burst (DC-CFB) analysis was developed recently [1] and is applied here to rigorously assess the loading efficiency of Ab-RBC and Ab-RBC + preparations.To that aim, the collected red and green fluorescence signals were time-gated according to their excitation pulses and fluorescence lifetimes, as shown in Figure S4a-d, where the red and green regions highlight the time windows for photon collection in the relevant spectral ranges.

S2-1-Background calculation
Before searching for the fluorescent bursts at each color, the background rates versus time were evaluated for each preparation at 50 s temporal intervals, as depicted in Figure S5.In summary, the evaluations reveal that the average background rates in antibody (dUTP)-loaded vesicles decreased approximately 4.6 (4.4) times in the red channel and 1.95 (2.75) times in the green channel after the additional cleaning process.In conclusion, these results underscore the efficacy of the supplementary cleaning step in eliminating more non-encapsulated cargoes from the final solution.
Remarkably, this cleaning process had a substantial impact on the green channel, leading to a fourfold reduction in its corresponding background rates.This reduction is primarily attributed to the removal of non-encapsulated antibody/dUTP cargoes.

S2-2-Assessment of total population of nanovesicles
Following the subtraction of time-dependent background rates from the red and green channels, the DC-CFB analytical method was employed to identify the red and green fluorescent bursts, corresponding to the entire nanovesicle population and the green-tagged cargo molecules, whether  Therefore, the determination of the total nanovesicle population involved identifying optimized key parameters for the DC-CFB analysis, including minimum photon number (M) and count rate (F) thresholds for both red and green bursts.As previously detailed, the optimal settings for   and   for the red bursts were established through a burst analysis protocol that systematically compared the red fluorescence results with independent AFM measurements to validate the extracted size distributions.
Meanwhile, there might be some differences between AFM and red fluorescence size distributions due to hydration effect in the latter one as it is conducted in physiological buffer solution.The outcomes of the total population assessment using the red fluorescence signal (  ) and their comparison with corresponding AFM size profiles are depicted in Figure S6, for two different preparations (RBC and RBC + ) and cargo molecules (Ab and dUTP).respectively.In d, h), the difference between the curves presented at c, g), respectively, serving as the basis for error minimization utilized in the optimization of burst data, as detailed in Ref.Moreover, after calibrating the brightness of green-tagged cargos inside the RBC and RBC + nanovesicles by comparing their lifetimes with the corresponding free cargos in solution as the brightness ratio, listed in Table S3 and shown in Figure S11 c-d.

, 40 Figure 1 .
Figure 1.a) Preparation steps of antibody loaded nanovesicles.Ab = antibody, RBC=red blood cell nanovesicles, RBC + = RBC with additional solution cleaning.b) Dual-color fluorescent staining scheme for the Ab cargo (Alexa488, green dye) and the nanovesicle (CellVue Claret, red dye).Atomic force microscopy (AFM): c) experimental setup and d) image of nanovesicles.Dualcolor fluorescence microscopy (DCFM): e) setup and f) typical time traces in detection for the red (vesicles) and the green (antibody molecules) signal channels.Coincident red and green temporal bursts denote Ab-loaded nanovesicles.

Figure 2 .
Figure 2. Size histograms of the whole populations of nanovesicles assessed through a-b) AFM and c-d) red burst analysis in fluorescence microscopy experiments, for RBC (plots a and c) and RBC + (plots b and d) sample preparations.
and essentially absent in the narrower AFM profiles of Fig. 2a-b.Furthermore, both AFM and fluorescence results consistently indicate no significant impact of the additional cleaning step (RBC vs RBC + ) on Ab-loaded sample preparations.The maximally populated nanovesicle radius derived from the fluorescence data analysis remains the same for both RBC and RBC + samples (  = 32 ) and this is equally true for their average radii (  ~ 60 ).

Figure 3
Figure 3 illustrates the result of further investigations into the sub-populations of loaded

Figure 3 .
Figure 3. Histogram of the number of Ab-loaded nanovesicles (  ) as a function of nanovesicle radius R, for: a) RBC and b) RBC + sample preparations.Size-distribution of the loading yield (), for Ab (solid lines) and dUTP (dashed lines) cargo molecules, for : c) RBC and d) RBC + preparations.

Finally, the
single-molecule resolving capability of fluorescence measurements, combined with further analyses and calibration experiments detailed in S2 and Ref. 35, afforded also statistical investigations on the number-normalized brightness per vesicle (  ), illustrated in Fig. 4a-b.

Figure 4 .
Figure 4. Histograms of a number-normalized brightness per vesicle (  ), for : a) RBC and b) RBC + sample preparations.The normalized lifetime histograms of Alexa488 dye bound to Ab-cargo molecules in three conditions such as a1: free in solution, a2: loaded into c) RBC nanovesicles, and a3: loaded into d) RBC + nanovesicles.

Figure 5 .
Figure 5. Profile of antibody-loaded nanocarriers: Two dimensional histograms of loaded nanovesicles versus their radius () and number-normalized brightness per vesicle (  ), obtained from dual-color coincident fluorescence burst assessments.The size and number of Ab-cargo molecules per vesicle for the loaded vesicles are depicted in the horizontal and vertical subplots, respectively, for a) RBC and b)RBC + preparations.It demonstrates that loaded nanovesicles are mostly populating at radius of 52 nm and maximum brightness-normalized number of 1.75 and 2.25 Abs per RBC and RBC + nanovesicles, respectively.

Figure S1 .
Figure S1.The schematic of the core processes associated with the loading of Ab-Alexa488 cargos into DRM nanovesicles derived from RBC-membrane, including the final preparation stages such as red fluorescence tagging of vesicles, highlighted in red box.
, the real pictures of the prepared samples in different steps are illustrated, including a) the top-loading of purified erythrocyte ghosts in 1% triton x-100 on density gradient and several formed sucrose buoyancies (10%, 24%, 30% and 40%), b) the separated erythrocyte ghosts on density gradient after being treated with 1% triton x-100, while the upper small band is detergent resistant membranes and the broader lower band is bulk erythrocyte membrane which are not floating up as the lighter DRMfraction, c) DRM-nanovesicles loaded with Ab-Alex488, and finally d) the DRM-nanovesicles stained with far red namely as Ab-RBC.

Figure S2 .
Figure S2.The real picture of the samples during each preparation steps encompassing a) the top-loading of purified erythrocyte ghosts in 1% triton x-100 on density gradient, b) the separated erythrocyte ghosts on density gradient after being treated with 1% triton x-100, while the upper small band is detergent resistant membranes and the broader lower band is bulk erythrocyte membrane that are not floating up as the lighter DRM-fraction, c)

Figure S4 .
Figure S4.The time resolved lifetime histograms of fluorescent green and red channels highlighting their respective lifetimes (Nanotime bin=16 ps) and time gating windows (shaded green and red areas) for: a-b) Abloaded and c-d) dUTP-loaded RBC and RBC + preparations.

Figure S5 .
Figure S5.Calibration measurements of the background rates (BG) for the green and red channels performed in four different experimental sessions, on: a-b) Ab-and c-d) dUTP-loaded nanovesicles for samples without (RBC, solid lines) and with (RBC + , line with markers) additional cleaning steps (see S1).
located inside or outside the RBC nanovesicles, respectively.Based on the Stokes-Einstein diffusion theory, each burst duration time (τburst) in the fluorescence time traces obtained from the experiments was converted into the radius () of the nanovesicles or molecules on single-vesicle and singlemolecule basis.This conversion was accomplished as  = 4   6µ 0 2   , where kB, T and µ represent the Boltzmann constant, the lab temperature, and the viscosity of the solvent (water), respectively.The S10 lateral radius of the confocal volume, denoted as w0, was measured in calibration experiments as  0 = 319  and  0 = 254  for the red and green laser spots, respectively.

Figure S6 .
Figure S6.Nanovesicle size distributions normalized to their peak value obtained by AFM (dash-dotted brown) and fluorescence red burst analyses (solid red) for: Ab-loaded a) RBC, b) RBC + , dUTP-loaded c) RBC and d) RBC + preparations.

Figure S8 .
Figure S8.Size histograms of loaded nanovesicles obtained from fluorescence experiments for a-c) dUTP-RBC and e-g) dUTP-RBC + preparations.In a, e), the histograms are derived from red fluorescence analyses, while in b, f), their result from coincident green analyses are provided, following the procedures and methodologies outlined in Ref. 1. Corresponding c, g) depict normalized histograms of the nanovesicle distributions, derived through coincident-red and coincident-green results, highlighted by solid red and dash-dotted green lines,

Figure S10 .
Figure S10.Size histograms retrieved from fluorescence measurements for: a-b) the loaded nanovesicle subpopulations (  ) and c -d) corresponding loading yield (()) profile (determined by dual color burst analysis), in the case of dUTP-loaded samples including RBC (a and c) and extra cleaned RBC + (b and d) preparations.

Figure S13 .
Figure S13.Two dimensional profiles of dUTP-loaded a) RBC and b) RBC+ nanovesicles, obtained from dualcolor coincident fluorescence burst analysis.The horizontal and vertical subplots depict the size and numbernormalized brightness per vesicle for dUTP-cargo molecules, respectively.It demonstrates that loaded nanovesicles are mostly populated at radius of 42 nm for RBC (48 nm for RBC+) and maximum number of 2.25 dUTPs per nanovesicle for both preparations (RBC and RBC + ).

Table I .
35e latter (see also SI) were processed at the same time and under identical experimental conditions, for a direct comparison with the Ab-loading cases and to provide a reference to previous literature.35Summary of the statistics of the RBC and RBC respectively.For a direct comparison, TableIreports the values retrieved by AFM and fluorescence measurements for both preparations subject to Ab-(rows 1-2) and dUTP-(rows 3-4) loading processes.+nanovesiclepopulations subject to the loading procedures of Fig.1a, with Ab or dUTP as cargo molecules, retrieved from AFM and red fluorescence burst analyses.  = nanovesicle radius at the peak of their size-distribution,   = average radius for each vesicle population,   = standard deviation of the vesicle radius.

Table II .
Summary    = average radius of the loaded vesicles,   = standard deviation of the vesicle radius.  = loading yield for R=   ,   = average loading yield.  and   are the maximum and average number of loaded cargos per vesicle (  or   ), respectively.
of the statistics of the loaded sub-populations of RBC and RBC + nanovesicles, with Ab and dUTP cargo molecules, retrieved from dual-color coincident fluorescence burst analyses.   = loaded nanovesicle radius at the peak of their size distribution, ~145 KDa) of antibodies compared to labeled dUTP molecules (~1 KDa).Another notable difference is observed in the dUTP-loaded vesicle distributions after the extra-cleaning process, evident from the data in S3 and TableII, indicating an increase by 6 nm in both    and    of RBC + versus RBC dUTP-loaded nanovesicles.This consistent shift to larger sizes also occurs in the statistics of the overall nanovesicle populations when comparing RBC and RBC + preparations in the dUTP case, highlighting the size-filtration effect of the original nanovesicle populations associated with the RBC + cleaning step, favoring slightly larger vesicles (R≥50 nm) that better match Ab-loaded vesicles.This justifies the observed changes affecting the dUTP-loaded and not the Ab-loaded vesicles, as well as the modification of the dUTP-loading yield results in RBC and RBC + preparations, apparent in Fig.3c-d(dashed lines) and in Table II (  and   in rows 3-4).Further analysis of the dual-color experimental data, explained in SI, enables size-resolved evaluations of loading yields, as depicted in Fig. 3c and 3d for RBC and RBC + samples, respectively.Consistent with previous discussions, the Ab-loading yield distribution, (), remains unaffected by the cleaning procedures, peaking at the same vesicle radius,    = 52 , for both RBC and RBC + preparations.The maximum loading yield,   = ( , depicting the total nanovesicle populations for the Ab-loading case, highlights a shift of the vesicle distributions toward larger sizes upon loading.This shift is quantified by comparing the values for   in TableIand    in TableII, revealing an increase of ∼20 nm in the peak radius of Ab-loaded compared to the whole nanovesicle populations.This size increase in the loaded nanovesicle population is also apparent in the values of the average radii of Ab-loaded (   , TableII) and whole (  , TableI) vesicle populations.In comparison to dUTP-loaded vesicles, the Ab-loaded vesicles exhibit approximately 10 nm-larger average and peak sizes, consistent with the larger size and molecular weight ( ), is also minimally affected, with values of 38% and 41% for RBC and RBC + samples, respectively (TableII).Average values of loading efficiencies and vesicle sizes exhibit similar trends, with   = 14% and    ~ 65 , respectively, regardless of the extra cleaning step in the Ab The work was supported by the program for biological pharmaceutics of the Swedish Innovation Agency (Vinnova grant no 2017-02999), and by the OQS Research Environment for Optical Quantum Sensing of the Swedish Research Council (VR grant no 2016-06122).K. G. gratefully acknowledges further support from the Knut and Alice Wallenberg Foundation through the Wallenberg Center for Quantum Technology (WACQT) and from the Swedish Research Council via grant VR 2018-04487.J. W. acknowledges further support from the Swedish Research Council via grant VR 2021-04556.

Table S1 .
1. Summary for the optimized parameters defined according to the data analysis parameters (F: the minimum photon-rate threshold and M: the minimum photon counts per burst) and protocols of Ref. 1 for the red and green fluorescence burst analyses of the four typologies of both cargo and vesicle preparations considered in the study (Ab-/dUTP cargoes and RBC/ RBC + preparations).{()} and 〈()〉 are the corresponding