Differences of the tumour cell glycocalyx affect binding of capsaicin-loaded chitosan nanocapsules

The glycocalyx regulates the interaction of mammalian cells with extracellular molecules, such as cytokines. However, it is unknown to which extend the glycocalyx of distinct cancer cells control the binding and uptake of nanoparticles. In the present study, exome sequencing data of cancer patients and analysis of distinct melanoma and bladder cancer cell lines suggested differences in cancer cell-exposed glycocalyx components such as heparan sulphate. Our data indicate that glycocalyx differences affected the binding of cationic chitosan nanocapsules (Chi-NCs). The pronounced glycocalyx of bladder cancer cells enhanced the internalisation of nanoencapsulated capsaicin. Consequently, capsaicin induced apoptosis in the cancer cells, but not in the less glycosylated benign urothelial cells. Moreover, we measured counterion condensation on highly negatively charged heparan sulphate chains. Counterion condensation triggered a cooperative binding of Chi-NCs, characterised by a weak binding rate at low Chi-NC doses and a strongly increased binding rate at high Chi-NC concentrations. Our results indicate that the glycocalyx of tumour cells controls the binding and biological activity of nanoparticles. This has to be considered for the design of tumour cell directed nanocarriers to improve the delivery of cytotoxic drugs. Differential nanoparticle binding may also be useful to discriminate tumour cells from healthy cells.


Results and discussion
Characterisation of the glycocalyx of different human cells. Exploration of public exome sequencing data of bladder cancer and melanoma tissue revealed that genes coding proteoglycans, enzymes of the heparan sulphate or hyaluronic acid biosynthesis and enzymes degrading heparan sulphate or hyaluronic acid are frequently mutated or affected by copy number alterations (CNAs). Figure 1 shows that the frequency of somatic mutations was higher in melanoma than in bladder cancer. In contrast, CNAs were more frequently detected in bladder cancer tissue and less often in melanoma. These data suggest that the glycocalyx of tumour cells of both entities is heterogeneous and varies between different patients and most likely even between individual cells within the same tumour tissue.
Next, we analysed the glycocalyx of urothelial (T24, UROtsa) and melanoma (BLM, MV3) cells of different origin and phenotype. T24 cells are derived from transitional bladder cancer, while UROtsa cells resemble the healthy urothelium [35][36][37] . The malignant melanoma cell line BLM has been isolated from lung metastasis 38,39 , www.nature.com/scientificreports/ whereas MV3 cells originated from lymph node metastasis 40 . Figure 2a shows the analysis of glycocalyx-related gene expression by quantitative real time PCR (qRT-PCR). Genes encoding enzymes involved in heparan sulphate synthesis, such as exostosin glycosyltransferase 1 (EXT1) and EXT2, were strongly expressed, as were those encoding the heparan sulphate proteoglycans syndecan 1 (SDC1) and SDC4. SDC2 was weakly expressed by MV3 cells. SDC3 was detected at low levels in T24, and both melanoma cells. In none of the analysed cell lines we detected the expressions of hyaluronan synthase 1 (HAS1). HAS2 was weakly expressed by T24, MV3 and BLM cells. HAS3 was found in UROtsa cells only. Hyaluronidase 1 (HYAL1) expression was below the detection limit in all tested cell lines, while HYAL2 was weakly expressed by T24, MV3 and BLM cells. (Fig. 2a, Supplementary  Table S1). In summary, qRT-PCR data suggest that in all tested cell lines heparan sulphate is more abundant than hyaluronic acid. However, the biosynthesis of glycosaminoglycans is a complex process. For example, heparan sulphate biosynthesis involves the action of 11 different enzymes, which cooperate with and antagonise each other, thus preventing a straight forward correlation between gene expression levels and the quantity of heparan sulphate found on the cell surface 41,42 . Therefore, to further characterise the glycocalyx ultrastructure, we applied stimulated emission depletion microscopy. Cell surface-exposed glycans were stained by ATTO 646N-conjugated wheat germ agglutinin (WGA). We found that the surface morphology was strongly dependent on the origin of the cell (Fig. 2b). While the surface of the urothelial cells (UROtsa and T24) was characterised by tube-like membrane protrusions (Fig. 2b, white arrow), both melanoma cell lines (BLM and MV3) showed bleb-like bulges  Supplementary Table S1. (b) Stimulated emission depletion microscopy images of the glycocalyx of T24, UROtsa, BLM and MV3 showing that the surface morphology was strongly dependent on the origin of the cell. T24 and UROtsa cells exposed tube-like cellular protrusions (white arrow), whereas BLM and MV3 cells produced bleb-like protrusions (white arrowhead). Scale bars = 1 µm. The shown inserts were placed in dark image areas to prevent overlapping of multiple fluorescence signals but to enable the resolution of single glycan spots.  Fig. 2b, white arrowhead). The formation of membrane protrusions has been mechanistically linked to the density of the mammalian cell glycocalyx 43,44 . Therefore, tube-like membrane folding, as found on the urothelial cells, may account for a high glycan density. In contrast, bleb-like structures may indicate a low glycan density. Proteoglycans are organised as clusters on the cell surface 45 in a tuft-like manner with interstitial gaps in the hundred-nanometer range 46,47 . In line with this concept, we detected single glycoprotein clusters on the surface of all cell lines as small dots (Fig. 2b), with an average spacing of ~ 200 nm (Fig. 2c) and a number of approximately 20 clusters per µm 2 (T24: 22 ± 2.9 clusters µm −2 ; UROtsa: 21 ± 1.4 clusters µm −2 ; BLM: 18 ± 2.5 clusters µm −2 ; MV3: 18 ± 1.4 clusters µm −2 ) 48 . Figure 2d shows a schematic drawing of transmembranous proteoglycans and the supposed presentation of glycosaminoglycans at the cellular surface.
Glycocalyx-dependent binding of Chi-NCs. The amount of surface-exposed glycans and in particular of heparan sulphate was determined by flow cytometry (Fig. 3, Supplementary Fig. S1). The applied WGA recognises sialic acids and N-acetylglucosamines, which is a principal component of heparan sulphate, hyaluronic acid and N-and O-linked glycans. For the detection of heparan sulphate only, we used the specific 10E4 monoclonal antibody.
WGA-and 10E4-binding to urothelial cells was of a comparable proportion suggesting that heparan sulphate is a major component of the urothelial glycocalyx and that in average T24 cells expose more heparan sulphate than UROtsa cells (Fig. 3a). Moreover, UROtsa cells showed a less homogeneous heparan sulphate exposure than T24 cells (Supplementary Fig. S1). In melanoma cells (Fig. 3b), we found comparable levels of WGA-binding, whereas the 10E4-binding indicates increased levels of heparan sulphate in BLM cells. The surface expression of glycan structures was further correlated to the ability of the cells to bind Chi-NCs. The Chi-NCs, schematically shown in Fig. 3c, have previously been characterised by our group using dynamic light scattering and transmission electron microscopy 25 . We re-evaluated their size and charge by Zetasizer measurements. They had a diameter of 157 ± 53 nm, a size polydispersity index of 0.2 ± 0.03 and a ζ-potential of + 49.5 ± 8.1 mV. Nanoparticle tracking analysis revealed a comparable diameter of 112 ± 30.2 nm.
Flow cytometry cannot discriminate between surface-bound and intracellular Chi-NCs. Because we aimed to measure the ability of the cells to bind Chi-NCs to their surface, we performed our experiments on ice slowing down Chi-NC endocytosis. To verify the cellular localisation of the Chi-NCs, we analysed the cell suspensions by fluorescence and phase-contrast microscopy prior to flow cytometry. Representative images, shown in Fig. 3d and e, confirmed that the majority of Chi-NCs were in close proximity to the cell membrane. In our quantitative analyses by flow cytometry, we assumed that the cell surface areas of all cells were of an equivalent magnitude. T24 cells bound more Chi-NCs than UROtsa cells which correlated with the elevated amounts of cell surface-exposed glycans (Fig. 3d). Also BLM cells, which expose more heparan sulphate than MV3 cells, bound more Chi-NCs suggesting again the contribution of heparan sulphate in particle binding (Fig. 3e). The discrepant binding of WGA and 10E4 to melanoma cells may indicate the presence of further glycans at the cell surface. Our qRT-PCR results suggest that in comparison to BLM cells, MV3 cells may carry more hyaluronic acid because they express higher levels of HAS2. (Fig. 2a).
Taken together, our data indicate that the tumour cell glycocalyx and especially heparan sulphate contribute to the binding of Chi-NCs to the surface of the here tested cancer cells. However, it is likely that also other glycosaminoglycans such as chondroitin sulphate or heavily glycosylated proteins such as mucins could trap cationic nanoparticles at cellular surfaces.
The bladder, and local neoplastic tissues within it, can be accessed by endoscopic surgery, and anti-tumour drugs could be administered via catheters to minimise systemic adverse effects. The instillation of nanoparticles in defined buffers or media may also improve nanoparticle stability and may prevent unintentional modifications such as the formation of protein coronas. The bladder thus appears to be an ideal target for nanotherapeutics. Therefore, we focused on the bladder cancer cell line T24 and the urothelial cell line UROtsa in the further course of our work.
Internalisation and processing of Chi-NCs. We loaded Chi-NCs with the model lipophilic cytotoxic drug capsaicin, which was previously shown to induce apoptosis in bladder cancer cells 34 . Capsaicin is insoluble in water, which limits its application as a free drug. Accordingly, we found that the addition of free capsaicin to the culture medium of urothelial cells resulted in its precipitation and the generation of crystals ( Supplementary  Fig. S2). In contrast, capsaicin can be loaded efficiently into the o/w emulsion of Chi-NCs due to its partitioning between the two phases 25 . We have previously shown that the association efficiency of capsaicin was above 90%. The release rate of capsaicin from the Chi-NCs in cell culture medium is low and only in the range of 0.3% per hour 25 .
In the next set of experiments, we applied structured illumination microscopy to measure Chi-NC binding and uptake by adherent T24 and UROtsa cells. Representative images are shown in Fig. 4a. After 2 h of Chi-NC treatment, most of the Chi-NCs are close to the upper surface of the cells. After 24 h of permanent exposure, Chi-NCs appeared to be partially accumulated in perinuclear regions, suggesting Chi-NC uptake via endocytosis followed by endosomal deposition 49,50 . We further investigated the time-dependent Chi-NC internalisation and the potential impact of different capsaicin concentrations (125, 250 and 500 µm capsaicin) on UROtsa and T24 cells by live-cell fluorescence microscopy (Fig. 4b). Representative videos of cells treated with 250 µm capsaicin are shown in Supplementary Video S1 and S2. After 5 and 10 h, cells treated with 250 or 500 µm encapsulated capsaicin formed vacuoles in a dose-dependent manner, indicating cellular stress and autophagy due to the intracellular engulfment of the cytotoxic drug 51,52 . Only few vacuoles were visible in cells treated with 125 µm capsaicin. For quantitative analysis, the formation of vacuoles was measured over time (  where N v is the number of vacuoles (experimentally measured in the images as an area fraction), N v , max is the maximum detectable number of vacuoles, n is an arbitrary factor, and k −1 is proportionally to the ability of cells to deposit capsaicin in vacuoles. Figure 4d shows that vacuoles formed more frequently and faster in T24 cells and UROtsa cells (right) either treated with RPMI-1640 medium (control) or treated with capsaicin-loaded Chi-NCs (= Chi-NC-Cap) at a concentration of 500 µm capsaicin (5.6 × 10 12 particles mL −1 ) and with Chi-NCs without capsaicin (5.6 × 10 12 particles mL −1 ). Data are presented as mean ± SD (n = 3). (f) Snapshots of live cell fluorescence microscopy images of T24 cells, 0, 2.5, 5 and 10 min after treatment with DiO-loaded Chi-NCs (5.6 × 10 12 particles mL −1 ) indicating the rapid binding of Chi-NCs to the cell surfaces. Scale bar = 50 µm.
Scientific Reports | (2020) 10:22443 | https://doi.org/10.1038/s41598-020-79882-y www.nature.com/scientificreports/ than in UROtsa cells at a capsaicin concentration of 250 µm. At the highest Chi-NC dose (500 µm capsaicin), T24 and UROtsa cells formed a comparable number of vacuoles ( Supplementary Fig. S3). Moreover, after a prolonged incubation time (24 h) and permanent exposure to the capsaicin-loaded Chi-NCs, both cell lines detached from the culture flask. This suggests that the high dose of capsaicin was lethal for T24 but also UROtsa cells and that the per se reduced Chi-NC internalisation by UROtsa cells was overridden. Because capsaicin can also promote cell migration 53,54 , the dynamic behaviour of the cells was analysed by calculating frame-to-frame differences in our videos. In comparison to untreated cells (control), cellular motilities of T24 and UROtsa cells were slightly higher for few minutes after the administration (time point 0 h) of Chi-NCs and Chi-NC-Cap (5.6 × 10 12 particles mL −1 = 500 µm capsaicin). After this period of increased cellular motility, movement activity of Chi-NC/Chi-NC-Cap treated cells decreases and falls slightly below the control level, until it returned to normal. Of note, changes of the cell layer dynamics appeared to be capsaicin independent because capsaicin-free and capsaicin-loaded Chi-NCs showed a comparable effect. This is also in line with the low release rate of capsaicin from the Chi-NCs suggesting only minor levels of free drug, which are insufficient to induce cell migration (5.6 × 10 12 particles mL −1 = 500 µm capsaicin, Fig. 4e). Live fluorescence microscopy showed that the Chi-NCs bound rapidly to the cells (Fig. 4f) and that Chi-NC binding was also associated with cellular detachment, which may coincide with the initially increased cellular motility (Supplementary Video S3).
In summary, our results suggest that Chi-NCs efficiently deliver capsaicin into urothelial cells and that the cytotoxic response is most likely dependent on intracellular vacuole formation.
Capsaicin-induced cytotoxicity. To confirm the fluorescence microscopy data, we used electrical cellsubstrate impedance sensing (ECIS) for real-time quantification of cell behaviour 55 . ECIS or similar methods have frequently been used to measure the loosening of cell-cell contacts 56 . We applied ECIS to measure Chi-NC binding to the cell surface and the formation of vacuoles upon particle endocytosis. To enable a reliable data interpretation, we cultivated T24 and UROtsa cells until they formed a confluent cell layer prior to the treatment with Chi-NCs. Representative microscopic images of cultivated cells are shown in Supplementary  Fig. S4a. Absolute impedance of both cell lines was of a comparable magnitude before treatment ( Supplementary  Fig. S4b). The measured electrical impedance of both urothelial cell lines decreased rapidly after treatment with 500 µm nanoencapsulated capsaicin pointing towards an early cell detachment due to cytotoxicity, preventing reasonable data interpretation (Supplementary Fig. S4c and d). Therefore, we excluded those data from our analysis. After treatment with 125 and 250 µm encapsulated capsaicin or equivalent amounts of empty Chi-NCs (1.4 × 10 12 , 2.8 × 10 12 ) particles mL −1 ), we observed an increase in impedance which most likely reflects the binding of Chi-NCs to the cell surface (Fig. 5a,b). This initial increase of impedance peaked after 30 min and then underwent exponential decay. The latter observation may reflect the clearance of Chi-NCs from the cell surface by endocytosis which subsequently resulted in vacuole formation 57 . The response was more pronounced in T24 cells than in UROtsa cells at both Chi-NC concentrations, agreeing with the higher Chi-NC binding and vacuole formation capacity of T24 cells. In principle, ECIS is also sensitive to cell migration which may also account for changes in the measured impedance 58 . However, we consider this effect in our experiment as less relevant, because our microscopic analysis, shown in Fig. 4e, suggested that Chi-NCs affected the migratory potential of urothelial cells only slightly.
We assume that the time-dependent change in the impedance Z reflects the number of surface bound Chi-NCs and the formation of vacuoles. Accordingly, we correlated the measured Z to the binding of Chi-NCs and to the number of formed vacuoles as shown in Eq. (2): where Z 0 is the normalised impedance before Chi-NC treatment (t = 0 h), Z max is the maximum detectable impedance, a −1 is a factor proportional to the binding of Chi-NCs, and N v is the number of vacuoles from Eq. (1). Representative fitting of our impedance data is shown in Supplementary Fig. S5.
To facilitate better comparison with our flow cytometry and microscopy data, we segregated the fitted results and plotted the supposed Chi-NC binding and N v against time (Fig. 5c,d). In addition, Fig. 5e shows factor a −1 as a measure for Chi-NC binding. The ability of cells to deposit capsaicin in vacuoles (k −1 ) is shown in Fig. 5f. T24 cells bound significantly more Chi-NCs than UROtsa cells. Moreover, and in good agreement with our microscopic analysis shown in Fig. 4d, ECIS indicated that T24 cells formed more vacuoles than UROtsa cells (Fig. 5f). This in turn suggests that more capsaicin has been delivered to the T24 cells. No vacuole formation was detected at 125 µm capsaicin because the data fit did not converge to a reasonable result ( Supplementary  Fig. S6). WST-1 assays confirmed a reduced cell viability of T24 cells which were treated with 250 µm encapsulated capsaicin (Fig. 5g). In comparison, UROtsa cell were only slightly affected. In further experiments, we confirmed the induction of apoptosis by probing western blots of cell lysates for the apoptotic marker cleaved caspase-3 ( Fig. 5h and Supplementary Fig. S7). In the presence of 250 µm encapsulated capsaicin, the level of cleaved caspase-3 increased strongly in T24 cells but remained close to the detection level in UROtsa cells. Prolonged exposure of the western blot membrane revealed low levels of cleaved caspase-3 in UROtsa cell treated with 250 µm encapsulated capsaicin ( Supplementary Fig. S8). There was no reasonable caspase-3 cleavage in either cell line in response to 125 µm encapsulated capsaicin.
Our data suggest that the enhanced binding of Chi-NCs to T24 cells supports the cytotoxic action of capsaicin, whereas the viability of UROtsa cells was not strongly affected upon Chi-NC binding. We cannot exclude that other cellular mechanisms such as the intracellular processing of capsaicin or the Chi-NCs may account for the measured effects. However, our data illustrate that glycocalyx-dependent differences in nanoparticle binding and internalisation may also affect the biological activities of nanoencapsulated drugs. Our results indicate that T24 cells, in comparison to UROtsa cells, exhibit a pronounced and aberrant glycosylation which increased Scientific Reports | (2020) 10:22443 | https://doi.org/10.1038/s41598-020-79882-y www.nature.com/scientificreports/   59 . In our own previous work, we showed that elevated heparan sulphate levels on melanoma cells reduce their ability to metastasize into the lung, whereas the formation of lymph node metastasis was promoted 15 . In bladder cancer, an altered expression and composition of glycosaminoglycans such as the oncofoetal chondroitin sulphate has been reported 21 . Taken together these data indicate that the glycocalyx of cancer cells can be altered regarding expression levels and structure. This also implicates an altered capability of cancer cells to interact with nanoparticles compared with benign cells. However whether cancer cells have an increased or decreased capacity to bind nanoparticles might depend on the tumour entity and in particular on tumour cell specific features. Therefore, detailed knowledge of the cancer cell glycocalyx is required to optimize the therapeutic application of nanoencapsulated drugs.
To further characterise the contribution of the glycocalyx to Chi-NC binding and to better understand its molecular basis we analysed the interaction between Chi-NCs and the urothelial cell lines in more detail.
Cell-specific glycocalyx tunes the electrostatic interaction with Chi-NCs. Our data shown in Fig. 3 suggest that heparan sulphate is involved in the Chi-NC binding to urothelial cells. To further underline the expected action of electrostatic forces 60 , we measured the charge density (ζ-potential) on urothelial cell surfaces. In line with their higher Chi-NC binding ability, T24 cells are significantly more negatively charged than UROtsa cells (Fig. 6a). Moreover, reduced cellular binding of negatively charged nanoemulsions (diameter = 135 ± 28 nm, ζ-potential = − 31.4 ± 4.83 mV, Fig. 6b 25,26 ) further confirmed that the interaction between Chi-NCs and the urothelial cells is driven by electrostatic forces (Fig. 6b). Electrostatic interactions have often been exploited for the intracellular delivery of drugs and nucleic acids through different polycationic nanocarrier systems [61][62][63] .
Seiler and colleges 21 showed that chondroitin sulphate is a valid therapeutic target on bladder cancer cells. Because chondroitin sulphate is, similar to heparan sulphate, highly negatively charged, it may also contribute to the binding of Chi-NCs to bladder cancer cells. Therefore, to further clarify the contribution of glycocalyx components such as heparan sulphate, chondroitin sulphate and hyaluronic acid, we treated T24 and UROtsa cells with heparinase, chondroitinase and hyaluronidase (Fig. 6c,d). In case of T24 cells, hyaluronidase and chondroitinase treatment had little effect on Chi-NC binding, whereas elimination of heparan sulphate by heparinase prevented Chi-NC binding efficiently. Interestingly, Chi-NC binding to UROtsa cells appeared to be more complex and dependent on hyaluronic acid, heparan and chondroitin sulphate. However, it is worth mentioning that the applied chondroitinase may also exhibit slight activity on hyaluronic acid 64 . UROtsa cells express also more heparanase (HPSE) than T24 cells (Fig. 2a). Accordingly, the intrinsic HPSE of UROtsa cells can partially remove heparan sulphate from the cellular surface also during chondroitinase or hyaluronidase treatment.

Heparan sulphate related counterion condensation affects Chi-NC binding. It is known from
polymer science that the mobility of counterions, and thus their charge screening effect, depends on the charge density of the proximate polyelectrolyte. Previous data suggest a strong charge screening effect along heparan sulphate chains through counterion condensation 6 . Whereas high counterion mobility and moderate charge screening occurs in polyelectrolytes with low charge densities, the counterion mobility approaches zero (counterion condensation) for highly-charged polyelectrolytes, resulting in strong charge screening. We assume that counterion condensation may also influence the electrostatic interaction between oppositely charged mamma- where b is the average distance between charges along the polyelectrolyte chain, e is the elementary charge, ε 0 is the vacuum permittivity, ε is the dielectric constant of the solvent (78.5 for water at 25 °C), k is the Boltzmann's constant, and T is the absolute temperature. The value of b is calculated as shown in Eq. (4): where L is the contour length of the polyelectrolyte and P is the number of charged groups. The counterion condensation theory predicts that condensation occurs when ξ > 1. Heparan sulphate can carry up to four charges per nanometre, which based on Eqs. (3) and (4) gives b = 0.25 nm and ξ = 2.9, supporting the occurrence of counterion condensation. Comparable electrostatic properties could be supposed for chondroitin sulphate. Hyaluronic acid is less negatively charged and most likely not affected by counterion condensation 65 .
In our preceding experiments, we applied high saturating Chi-NC concentrations. Therefore, we expect that the Chi-NC binding was mainly controlled by the total number of available binding sites at the cell surface (Fig. 3). In addition, high concentrations of Chi-NCs ensured also that sufficient capsaicin was delivered into the cells to induce apoptosis (Fig. 5). To analyse the potential impact of charge screening on the binding of Chi-NCs to mammalian cells, we treated both urothelial cell lines with different and lower concentrations of Chi-NCs (1.12 × 10 11 to 1.78 × 10 12 particles mL −1 ). Figure 7a shows the dose-dependent Chi-NC binding to UROtsa and T24 cells. To prevent misinterpretations due to the differences in Chi-NC binding capacity of T24 and UROtsa cells (Fig. 3d), we normalised the maximal measured mean fluorescence of T24 and also of UROtsa cells to 100%. At low Chi-NC concentrations and thus below saturation and independent of other limiting factors, such as differences in cell size, particle binding increases linearly with increasing amounts of Chi-NCs. However, the dose-dependent increment of particle binding was 2.5 times steeper in the UROtsa cell group than in the T24 cell group (UROtsa: 17.0 × 10 -11 (%(Chi-NC binding) ml n −1 ); T24: 6.6 × 10 -11 (%(Chi-NC binding) ml n −1 )). Moreover, the binding curve of the T24 cells suggests a cooperative binding behaviour, with a weak binding rate at low Chi-NC doses and an increasing rate towards higher Chi-NC concentrations before saturation is reached. Following the idea of counterion condensation, the weak interaction between T24 cells and Chi-NCs at low doses may reflect the screening of negative charges on the heparan sulphate chains by counterions, which results in less potential binding sites for Chi-NCs.
Recently, counterion condensation on heparin was shown to increase with the heparin chain length, also suggesting an increase in counterion condensation on cells with high levels of heparan sulphate, such as T24 cells 6 . Interestingly and in contrast to T24 cells, Fig. 6b indicated that Chi-NC binding to UROtsa cells was at least partially dependent on hyaluronic acid which is not sufficiently charged to provoke counterion condensation 65 . This may also explain the non-cooperative binding of Chi-NCs to UROtsa cells (Fig. 7a). In further experiments, we analysed Chi-NC binding in a salt-free sucrose buffer, anticipating the promotion of electrostatic interactions and the abolishment of the counterion condensation effect ( Figure S9). Indeed, under these conditions, the cooperative binding of Chi-NCs to T24 cells was attenuated indicating less counterion condensation. Because the absence of salt is stressful for mammalian cells, we aimed to confirm the impact of the supposed counterion condensation by additional experiments. Accordingly, we repeated our flow cytometry experiments with T24 cells treated with heparinase. We assumed that a thinner glycocalyx with shorter heparan sulphate chains exhibit less counterion condensation and in turn an improved Chi-NC binding at low particle doses. Although the cooperativity of the cell-particle interaction was not completely abolished, normalised particle binding rate to T24 cells with chopped heparan sulphate chains was increased in comparison to control T24 cells (Fig. 7b). Further in silico modelling of our data enabled us to estimate the initial effective charge, occurring at the cellular surface prior to the addition of Chi-NCs (Fig. 7b, dashed lines). Our model assumed a reduced number of binding sites for Chi-NCs below a critical nanoparticle concentration due to counterion condensation. The related decreased probability of binding events attenuated the number of cell-bound particles. Above the critical Chi-NC concentration, the release of tightly bound counterions increased the number of binding sites and thus the probability of binding events. We further translated the assumed initial effective charge to a corresponding initial charge density (ξ 0 ) as previously reported 6 . Accordingly, the calculated ξ 0 value of the control T24 was 4.2, thus indicating strong counterion condensation, whereas upon heparinase-treatment ξ 0 was assumed to be 1.5, suggesting only weak condensation. Moreover, our data enabled us to determine the critical amount of Chi-NCs, c*, required for the transition from the condensation regime at low particle concentrations to non-compromised counterion mobility at higher Chi-NC doses (Fig. 7b). For transition in the control T24 cells, 2.6 × 10 11 Chi-NCs ml −1 were required; whereas in T24 cells that underwent heparan sulphate degradation the transition occurred already at c* = 1.2 × 10 11 Chi-NCs ml −1 . In summary, we postulate that the interaction between mammalian cells with an extended glycocalyx and cationic nanoparticles is diminished at low doses (Fig. 7c, left), while disruption of the counterion condensation regime above a critical amount of nanoparticles releases screened binding sites at the cellular surface which promotes further nanoparticle binding (Fig. 7c, right). By determine c*, we were able to calculate a length of 230 ± 36 or 109 ± 16 nm for the heparan sulphate chains covering non-treated or The theory of counterion condensation is applicable to several polyions but may not completely describe the effects we measured at the cellular surface 66 . Detailed molecular dynamics simulations may offer an interesting way to understand better the occurrence and relevance of counterion condensation at the glycocalyx of mammalian cells 67 . and UROtsa cells at increasing Chi-NC concentrations (1.12 × 10 11 to 1.78 × 10 12 particles mL −1 ) measured by flow cytometry. Data are presented as mean ± SD (n = 4). Because UROtsa cells bound significantly less Chi-NCs than T24 cells (Fig. 3d), we normalised the maximum of detectable fluorescence of both cells to 100%. Dashes lines indicate the initial dose-dependent increment of the measured fluorescence. (b) Chi-NC binding to control (red circles) or heparinase-treated (blue circles) T24 cells. The grey dashed line indicates in silico simulations of Chi-NC binding to cell surfaces in the absence of counterion condensation (initial charge density (ξ 0 ) < 1), the blue dashed line shows simulation of Chi-NC binding at ξ 0 = 1.5 suggesting weak counterion condensation and the red dashed line shows simulation of Chi-NC binding at ξ 0 = 4.2, representing strong counterion condensation. Chi-NC concentrations = 4.97 × 10 10 to 1.27 × 10 12 particles mL −1 ). (c) Schematic illustration of the proposed counterion condensation effect regulating Chi-NC binding to the negatively charged heparan sulphate chains. Prior to the treatment with Chi-NCs, counterions are tightly bound to the heparan sulphate chains due to counterion condensation (left). The initial charge density (ξ 0 > 1) results in the strong screening of negative charges by counterions, which lowers the effective charge. Cooperative binding of Chi-NCs to heparan sulphate neutralises some of the negative charges leading to a reduced charge density (ξ < 1), no counterion condensation and thus to an increased effective charge that promotes further Chi-NCs binding (right).

Conclusion
In the past, the glycocalyx of mammalian cells has been investigated extensively to determine its role in cell migration and the perception of extracellular signals, but knowledge on its role for the interaction of cells with nanoparticles is scarce. In line with the public databases, our results suggest that the glycocalyx of different tumour cells of distinct entities are highly heterogeneous, which appear to be further translated to different nanoparticle binding capacities. However, we found that particle binding and the composition as well as the extent of the glycocalyx are not related in a simple manner. The contribution of different glycocalyx components and charge screening effects through counterion condensation must be taken into account. This work focused on the urothelial glycocalyx because the bladder is a hollow organ amenable for direct and local administration of nanoparticles, thus circumventing the limitations of systemic delivery such as the formation of a protein corona and nanoparticle clearance. However, the crosstalk between the glycocalyx and nanoparticles appears to be a common feature affecting the interactions between nanoparticles and all cancer cells. Here, we present a quantitative approach that enables to estimate the critical number of nanoparticles needed to maximise their binding to the glycocalyx of urothelial cells, thus promoting a localised drug delivery. This knowledge may be realised in future applications aiming to the rational dosage of nanoparticles for therapy or imaging purposes. Advanced knowledge of the acting electrostatic forces appears to be required to improve nanoparticle-mediated drug targeting and to prevent the generation of drug resistances during therapy.

Methods
Additional methods are described in the supplemental methods section (Supplementary Information).
Cell culture. The human bladder cancer cell line T24 (derived from urinary bladder carcinoma 68,69 ), the benign immortalised urothelial UROtsa cell line (kindly provided by Dr. Phillip Erben, Heidelberg University; derived from the urothelium lining the ureter of a 12-year-old girl [35][36][37]70 ) as well as the human metastatic melanoma cell lines MV3 (developed from lymph node metastases 40 ), and BLM (subline of BRO, derived from a primary human melanoma, isolated from lung metastasis after subcutaneous inoculation of nude mice with BRO cells 38,39 ) were cultured in RPMI-1640 medium (Sigma-Aldrich, Darmstadt, Germany) supplemented with 10% fetal bovine serum (FBS), 1% l-glutamine (Biochrom, Berlin, Germany) and 1% penicillin/streptomycin (Biochrom) as described previously 71 . The cells were incubated in a humid atmosphere at 37 °C with a 5% CO 2 atmosphere until they achieved a confluency of 90%.  25 . Unloaded Chi-NCs were prepared by replacing the capsaicin solution with ethanol. The organic phase was made up to 10 mL with ethanol and then immediately poured over 20 mL of an aqueous chitosan solution (0.5 mg mL −1 dissolved in water with a 5% stoichiometric excess of 1 M HCl) containing chitosan HMC 70/5 with a degree of acetylation of 32% (as determined by 1 H NMR) and a molecular weight of 5.01 × 10 4 g mol −1 (Heppe Medical Chitosan, Halle, Germany), the same material as in our previous studies 26 . The Chi-NCs formed spontaneously due to solvent displacement and Marangoni effects 72 . Finally, the ethanol and part of the water were evaporated at 40 °C under vacuum in a R210 Rotavapor (Büchi Labortechnik, Essen, Germany) and the volume of the formulations was reduced to 6 mL 26 . The nanoemulsions were prepared using the same procedure but without including chitosan. A Zetasizer NanoZS (Malvern Instruments, Malvern, UK) fitted with a red laser (λ = 632.8 nm) was used to measure the size and ζ-potential of the nanoformulations 25 . Diameter of the Chi-NCs was verified by nanoparticle tracking analysis (NanoSight LM10, NanoSight, UK). The concentration of free (non-cell bound) fluorescent Chi-NCs was determined by tracking the DiO-loaded nanoparticles on an Axiovert 200 fluorescence microscope (Zeiss, Oberkochen, Germany) equipped with a 40 × oil objective with a numerical aperture of 1.4 (Zeiss) as previously described 73 . Images of 256 × 256 pixels were recorded using a water-cooled iXon3 camera (Andor Technology, Belfast, UK) at a rate of 10 frames per second and an EMCCD gain of 300. Number of cell bound Chi-NCs was calculated by subtracting the number of free Chi-NCs from the total number of Chi-NCs.

Preparation and characterisation of nanoformulations.
Stimulated emission depletion microscopy. Stimulated emission depletion microscopy and confocal microscopy were carried out in sequential line scanning mode using an Abberior stimulated emission depletion expert line microscope as previously reported 74 . Briefly, cells were seeded on glass coverslips and cultivated until they reached confluency. After fixation of the cells with 2% paraformaldehyde (Electron Microscopy Sciences) in HEPES buffered Ringer's solution, cell surface glycans were stained with WGA (Sigma-Aldrich) conjugated with an ATTO 646 N protein labelling kit (JenaBioscience, Jena, Germany). The stimulated emission depletion microscopy setup was based on a Nikon Ti-E microscope body for the excitation and detection of the fluorescence signal through a 60 × (numerical aperture 1.4) P-Apo oil immersion objective. A pulsed laser was used for excitation at 640 nm and a near-infrared pulsed laser (775 nm) for depletion. The detected signal was directed through a variable-sized pinhole (set to match 1 Airy at 640 nm) and detected using novel state-of-the-art avalanche photodiode APDs with appropriate filter settings for Cy5 (615-755 nm). Images were recorded with a dwell time of 3 µs, and the voxel size was set to 20 × 20 × 150 nm. Images were acquired in time-gating mode with  Fig. 3d,e, 6b and Supplementary Fig. S1, to 8.96 × 10 11 Chi-NCs mL −1 in the experiments shown in Fig. 6c,d, Fig. 7a  ECIS. The impedance of UROtsa and T24 cells was measured as previously described 75  Cell surface charge measurements. The ζ-potential of the UROtsa and T24 cells was measured as previously described 26 . Cells were cultivated until they reached 90% confluency, detached with trypsin and resuspended at a concentration of 100,000 cells mL −1 in isotonic, salt-free 30 mM HEPES buffer supplemented with 300 mM sucrose (pH 7.4) and containing different volume fractions of Chi-NCs. The ζ-potential was measured by mixed laser Doppler velocimetry and phase analysis light scattering (M3-PALS). Particle size and derived count rate were determined by dynamic light scattering with non-invasive back scattering (DLS-NIBS) at a measurement angle of 173°. The autocorrelation functions were matched to the default non-negative least squares (NNLS) fit to calculate the intensity size distribution plots and thus evaluate the Z-average diameter. A Malvern Zetasizer NanoZS fitted with a red laser (λ = 632.8 nm) was used for both methods and Zetasizer v7.12 was used to acquire and evaluate the data.
Scientific Reports | (2020) 10:22443 | https://doi.org/10.1038/s41598-020-79882-y www.nature.com/scientificreports/ The applied imageJ macro for automatized image analysis was self-made and can be obtained from the authors upon request. Briefly, phase-contrast images (8-bit format) were converted into a binary image that shows bright regions above a critical threshold in black and regions below a critical threshold value in white. Threshold values were set automatically using the "mean dark" threshold setting. From consecutive binary images, standard deviation was calculated using the imageJ "Z project". Standard deviation images of one experiment were stacked together and mean brightness of every image was calculated. Change of the mean brightness was considered to be proportional with the migratory potential of cells within the cell layer. To analyse the formation of vacuoles, a similar approach was used ( Fig. 4c and Figure Supplementary S3). Briefly, consecutive fluorescence images were loaded into imageJ and areas with specific fluorescence intensity range were defined using the auto threshold setting "triangle dark". The area of vacuoles was determined using the measurement function of imageJ. Data were further normalised to the average number of cells per field of view.

Statistical analysis.
Results are expressed as mean ± SD. Student's t-test was used to test for significant differences between the cell lines. P values below 0.05 were considered statistically significant.