Interactions of chlorophyll-derived photosensitizers with human serum albumin are determined by the central metal ion

Abstract Two structurally similar derivatives of chlorophyll a, chlorophyllide a (Chlide) and zinc-pheophorbide a (Zn-Pheide), differing only in central metal ion (Mg2+ or Zn2+, respectively) substituting the tetrapyrrole ring, were investigated with regard to their binding to human serum albumin (HSA). Chlide and Zn-Pheide are very promising photosensitizers with potential application in photodynamic therapy, therefore it is desirable to investigate their interactions with serum proteins. The studies included absorption and steady-state fluorescence spectroscopy, as well as molecular docking. It was found that both investigated compounds form complexes with HSA. Experimental data revealed two classes of binding sites for each compound. The affinities (Ka) for the first class were in the range of 105 and 106 M−1 for Chlide and Zn-Pheide, respectively, while the second class was characterized by the affinities of the order of 104 M−1 for both derivatives. Molecular docking simulations together with displacement studies revealed that the primary binding site of the studied compounds is the heme site, localized in the subdomain IB, however the best characterized binding sites of HSA, namely the Sudlow’s sites I and II are also involved. The interactions between the derivatives of chlorophyll and HSA were found to be predominantly hydrophobic and to a lesser extent hydrogen bonding. Our results demonstrate that the centrally bound metal ion determines both the affinity and mode of binding to HSA, which may be a feature differentiating these compounds in terms of their pharmacokinetics. Communicated by Ramaswamy H. Sarma


Introduction
Chlorophyllide a (Chlide) and zinc-pheophorbide a (Zn-Pheide) are semi-synthetic derivatives of chlorophyll a with potential application in photodynamic therapy. They possess several features of ideal photosensitizers (PSs), including strong light absorption in the part of visible spectrum, coinciding with the therapeutic window of human tissue, and a high efficiency of reactive oxygen species generation. Both, in vitro and in vivo studies have demonstrated a high photodynamic potential of Chlide and Zn-Pheide, making them good candidates for clinical use (Jakubowska et al., 2013;Szczygieł et al., 2014). Since the application of the PSs involves their direct administration into the bloodstream, the first issue to be taken into account when analyzing their pharmacokinetics is the interaction with the components of circulating blood. This involves binding to proteins, lipids and lipoproteins as well as to cell membranes of erythrocytes and leukocytes. Chlide and Zn-Pheide are relatively well water-soluble molecules, therefore they are expected to be transported by albumins and globulins rather than by lipoproteins and cell membranes (Mazi ere et al., 1991). And since albumin constitutes a bulk of serum proteins, it is predicted to be the main binder of the PSs, acting right after their application. In our previous studies, we observed a significant difference between Chlide and Zn-Pheide regarding their binding to bovine serum albumin (Szafraniec & Fiedor, 2021). This prompted us to thoroughly investigate the interaction of these compounds with HSA, which is of crucial importance from the point of view of their potential application as intravenously administered drugs. Binding to HSA is likely to significantly affect the pharmacokinetics and pharmacodynamics of the PSs and, as a consequence, an ultimate photodynamic effect resulting from their application. It can also attenuate the undesirable effect of dark cytotoxicity, which can occur at high doses of the PSs, and regulate the bioavailability of other drugs present in the organism simultaneously with the PSs.
Human serum albumin consists of a single chain with three structurally similar domains (I, II and III), each containing two subdomains, A and B (Carter et al., 1989). The subdomains IIA and IIIA are recognized as the principal sites of ligand binding (Sudlow et al., 1975;Sudlow et al., 1976;Kragh-Hansen et al., 2002). HSA is equipped with one Trp residue (Trp214), located in subdomain IIA, which, being its dominant intrinsic fluorophore, is often used in the association studies of HSA with endo-and exogenous ligands by fluorescence spectroscopy techniques (Lakowicz, 2006;Roufegarinejad et al., 2019). In the present study, spectroscopic analyses on the binding of chlorophyll-derived PSs to HSA were performed together with molecular docking studies in order to determine their binding affinities and binding sites. Absorption as well as fluorescence enhancement and quenching studies revealed the formation of complexes between the PSs and HSA. Additionally, fluorescence quenching and displacement studies showed the existence of at least two binding sites for the investigated compounds, which was further confirmed by molecular docking. The obtained comprehensive data were applied to explain the binding mechanism between chlorophyll derivatives and HSA, which may contribute to a better understanding of their transportation, distribution and metabolism and as a consequence their efficacy as PSs applied in vivo.

Materials
Human serum albumin (fatty acid free) was obtained from PAN-Biotech (Aidenbach, Germany). The protein was dissolved daily in PBS (10 mM phosphate, 140 mM NaCl, 2.68 mM KCl, pH 7.4). Chlorophyllide a was kindly provided by Prof. Leszek Fiedor and Dr. Małgorzata Szczygieł (Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krak ow, Poland). Zn-Pheide was obtained from pheophorbide a (Cayman Chemical, Ann Arbor, MI, USA) by direct metalation with zinc acetate, as described previously (Szczygieł et al., 2008). The purity of both compounds determined by HPLC analysis was at least 96%. The concentrations of the PSs were determined spectrophotometrically in their ethanol solutions, using the extinction coefficient at the Q y band 71,500 M À1 cm À1 . For each experiment, fresh solutions of the PSs were prepared by their suspension in appropriate volumes of DMSO. The structure of the PSs is shown in Figure 1.

Absorption measurements
Absorption spectra in the ultraviolet-visible region (UV-Vis) were obtained using an Evolution 201 spectrophotometer (Thermo Scientific, Waltham, MA, USA) equipped with xenon flash lamp. The spectra were recorded in the range 230 À 800 nm at room temperature, using a 1 cm pathlength quartz cuvette. The concentrations of HSA and PSs were 50 and 25 mM, respectively. The spectra were recorded for the protein and ligands separately, as well as for their complexes. The measurements were carried out in PBS (pH 7.4) at three concentrations of DMSO: 1%, 5% and 15%. In each measurement, the reference sample was the dissolution medium: PBS with appropriate concentration of DMSO.

Fluorescence measurements
Steady-state fluorescence spectra were recorded using a Jasco FP-8500 spectrofluorometer (Jasco, Pfungstadt, Germany) equipped with a 150 W xenon lamp. The measurements were performed in a Hellma fluorescence cuvette with 100 mL chamber volume. Each sample was incubated for 10 minutes at desired temperature to reach equilibrium before the measurement. The temperature was controlled using an ETC-815 watercooled Peltier thermostatic cell holder (Jasco). In all measurements, the excitation and emission slits were set to 5 and 10 nm, respectively. Each spectrum was measured twice, using a fresh aliquot of the same sample, in order to avoid a potential photodamage of HSA resulting from a PS excitation. Two types of fluorometric experiments were performed: a. Fluorescence enhancement of 1 mM PSs at increasing concentrations of HSA (0, 0.125, 0.25, 05, 1, 2, 5, 10, 25, 50, 100, 250 and 500 mM), with excitation at 430 nm and emission collection between 600 and 800 nm, at 310 K.
In these experiments the samples were prepared in PBS, pH 7.4 containing 1% DMSO. b. HSA (1 mM) fluorescence quenching after excitation at two different wavelengths, 280 nm (for Trp, Tyr and Phe residues) and 295 (for Trp only), and emission collection between 305 and 500 nm or 320 and 500 nm for these two excitation wavelengths, respectively. The PSs were added to HSA at concentrations of 0, 0.5, 1, 2.5, 5, 10, 25, 50 and 100 mM and the measurements were performed at two temperatures, 298 and 310 K. The final concentration of DMSO was 5%. The recorded fluorescence intensities were corrected for the inner filter effect, according to the equation:  (1) where F corr is the corrected fluorescence intensity, F obs is the measured fluorescence intensity, and A ex and A em are the absorbances of the measured solutions at the excitation and emission wavelengths, respectively (Lakowicz, 2006). For the calculation of dissociation constants, the fluorescence maxima corresponding to the bound fractions of the PSs recorded in the PSs fluorescence enhancement study were plotted against HSA concentration and fitted with two-site binding curve (birectangular hyperbola) in the Origin 2021 (Origin, Version, 2021).

Synchronous fluorescence measurements
Synchronous fluorescence spectra of HSA in the presence or absence of the PSs were collected using a FS5 spectrofluorometer (Edinburgh Instruments Ltd., Livingston, UK) equipped with a 150 W xenon lamp. The excitation and emission slits were set at 3 nm. The interval between the excitation end emission wavelength was set at 15 or 60 nm to observe the spectral behavior of Tyr and Trp residues, respectively. The measurements were performed in 1 cm quartz cell in PBS pH 7.4, at 310 K, maintained by a SC-20 water bath regulated thermostatic sample holder (Edinburgh Instruments Ltd.). In each assay, HSA (1 mM) was titrated with a PS at concentrations of 5, 10, 25 and 50 mM. The obtained fluorescence spectra were corrected for the inner filter effect using the Equation (1).

Displacement studies
Competitive binding studies were performed using three different site probes: hemin, warfarin and ibuprofen (all from Pol-Aura, Poland) for sites IB, IIA and IIIA, respectively. The concentrations of HSA (100 mM) and PS (1 mM) were kept constant and various concentrations of the site probes (0.5, 1, 2, 5 and 10 mM) were added to the system. After every addition of a probe, the sample was incubated for 10 min at 310 K to reach equilibrium. Fluorescence measurements were carried out using a FS5 spectrofluorometer (Edinburgh Instruments Ltd.) using 430 nm excitation wavelength and 600 À 800 nm emission range. The excitation and emission slit widths were set at 2 nm. The measurements were performed in PBS pH 7.4, at 310 K.

Molecular docking
The three-dimensional coordinates of HSA (PDB ID: 1AO6) were downloaded from the RSCB Protein Data Bank. Only the monomer chain A was retained and solvent molecules were deleted. ChemSketch 2020.1.2 software was used to draw the structures and pre-optimize them in 3 dimensions (ACD/ChemSketch, version 2.1, 2020). Further 3D structure optimization, as well as the calculation of partial (Mulliken) charges were performed in the ORCA 4.2 software using the BP86 functional with ZORA relativistic approximation and def2-SVP SARC/J basis set (Neese et al., 2020). The calculated charges were manually introduced into the pdbqt files. For each PS, eight active torsions in the side chains were included, and the tetrapyrrole ring was considered rigid. The protonation state corresponding to pH of 7.4 was set for HSA using OpenBabel 3.0.0 and Kollman charges were added to it in the AutoDock 4.2 (Morris et al., 2009;O'Boyle et al., 2011). A grid box covering the amino acid residues specific either to Sudlow's I, Sudlow's II or heme site was generated using the AutoGrid 4. Docking simulations were carried out in the AutoDock 4.2 using the Lamarckian Genetic Algorithm with at least 100 iterations for every binding site. Other AutoDock parameters were used with default values. Visualization of docking results was performed in the PyMOL and BIOVIA Discovery Studio Visualizer using the conformations with the lowest binding free energy (BIOVIA, 2021; Schr€ odinger, LLC., 2010). The association constants (K a ) for protein À ligand interactions were calculated from the obtained free energy change (DG) of docking using the equation:  where R (8.314 J mol À1 K À1 ) is the gas constant and T is the absolute temperature.

Results and discussion
The absorption spectra of the PSs in the presence of HSA differed significantly from the corresponding spectra of free PSs, which indicates the formation of complexes ( Figure 2). The increase in absorbance was connected with the blue shift of Q y maxima for both Chlide and Zn-Pheide, while in the Soret band blue shift was observed for Chlide and red shift for Zn-Pheide. Importantly, the increase in absorption maxima observed after the addition of HSA was larger for Zn-Pheide than for Chlide. Thus, in the Soret band we observed the increase of 0.27 for Chlide and 0.53 for Zn-Pheide. In Q y bands, in turn, the increase for Chlide was    equal to 0.12, while for Zn-Pheide 0.39. Taking into account that DMSO concentrations below 20% do not influence the structure of HSA, we measured the absorption spectra at three concentrations of DMSO: 1%, 5% and 15% (Batista et al., 2014). We observed that the absorption spectra of PS-HSA complexes measured at various DMSO concentrations were exactly the same, though the free PSs showed a progressive disaggregation with increasing DMSO concentration (data not shown). This indicates that HSA causes the maximal possible disaggregation of the PSs already at 1% of DMSO, and further increase in DMSO concentration is not necessary to increase the efficiency of binding. The subsequent fluorescence enhancement studies were therefore performed at 1% DMSO. The fluorescence of both Chlide and Zn-Pheide was found to substantially increase with increasing concentration of HSA, which can be explained by the fact that HSA acted as a solubilizer for PSs aggregates, as these compounds, though relatively hydrophilic, are not completely soluble in aqueous media (Figure 3, insets). In the case of Chlide, apart from the increase in fluorescence signal, about 6 nm blue shift of fluorescence maximum was observed (Figure 3(A), inset). This effect is most likely due to the loss of axially coordinated water molecules and a consequent change in the conformation of the tetrapyrrole ring (Bonnett, 2003;Fiedor et al., 2008;Szczygieł et al., 2008). No spectral shift was observed for Zn-Pheide (Figure 3(B), inset), probably due to stronger chelation of Zn 2þ , ion and different coordination properties (Hartwich et al., 1998;Szczygieł et al., 2008).
Since both PSs show a significant fluorescence in aqueous media even in their unbound state, to calculate their dissociation constants, the signal from their albumin-bound fraction had to be separated from that of free PS fraction. We therefore performed a correction of the obtained fluorescence signal which is fully described in Supplementary materials. To obtain the binding curves, the calculated bound PS fraction was plotted against corresponding HSA concentration (Figure 3). We observed that the obtained binding curves are much better fitted with two-binding site model than with a single binding site model (R 2 ¼ 0.95 vs. 0.99, respectively). Further increasing the number of binding sites did not improve the fit, but increased the inaccuracy of the determined coefficients. The dissociation constants determined on the basis of the fluorescence enhancement study are presented in Table 1.
To further investigate the interactions between HSA and the PSs, we measured the intrinsic HSA fluorescence before and after the addition of the ligands at variable concentrations. In HSA there are three types of fluorescent residues: tryptophan (Trp), phenylalanine (Phe) and tyrosine (Tyr), of which the strongest signal is obtained from Trp, due to its higher molar absorptivity and intrinsic fluorescence quantum yield (Jahanban-Esfahlan et al., 2017, 2021. The contribution of Phe residues to the intrinsic fluorescence of protein is negligible by virtue of its low absorptivity and a very low quantum yield (Lakowicz, 2006;Roufegarinejad et al., 2019). In the structure of HSA, there is only one Trp residue (Trp214), localized in its subdomain IIA, while Tyr residues are dispersed within all three domains (Hosainzadeh et al., 2012). In order to determine the participation of Trp214 and other fluorescent residues in the formation of the PS-HSA complexes, we performed the quenching experiments using two excitation wavelengths, k ex ¼ 280 nm for all fluorescent residues and k ex ¼ 295 nm for Trp only. The maximum of free HSA fluorescence, irrespectively of the excitation wavelength, was recorded at 363 or 364 nm. At very low concentrations of the PSs (up to 0.5 mM) we observed a slight increase in HSA fluorescence signal and the quenching started from 1 mM of the PSs (Figure 4). This biphasic effect is probably due to aggregation of HSA in the presence of DMSO and its disaggregation upon binding of ligands.
The quenching of HSA fluorescence can result from the transfer of energy between its fluorescent residues and the chromophores present in the PSs molecules. Such a transfer is possible when the distance between the donor and acceptor does not exceed 10 nm, which can result from complex formation (Ghisaidoobe & Chung, 2014). The quenching mechanism between the PSs and HSA can be described by the Stern-Volmer equation: where F and F 0 are the fluorescence emission intensities in the presence and absence of the quencher, respectively, K SV is the Stern-Volmer constant and [Q] is the concentration of quencher (Lakowicz, 2006). The Stern-Volmer quenching constant K SV is given by K SV ¼ k q Â s 0 , where k q is the bimolecular quenching constant (or the efficiency of quenching), and s 0 is the lifetime of the fluorophore in the absence of quencher. The value of s 0 for biopolymers is reported to be 10 À8 s (Cui et al., 2009). The quenching described by the Stern-Volmer equation may be both static and dynamic. It is considered that the threshold value of k q below which the quenching is predominantly diffusion-controlled equals to 2 Â 10 10 M À1 s À1 (Ge et al., 2010). For the measurements of quenching performed after the excitation at 280 nm, we obtained non-linear, downward facing Stern-Volmer plots in the case of both PSs ( Figure 5, left panel). This shape is characteristic for fractional accessibility of fluorophores to a quencher (Lakowicz, 2006). Corresponding plots obtained after the excitation at 295 nm were linear but biphasic, which shows that two binding classes of binding sites are responsible for the quenching ( Figure 5, right panel). These results show that a part of Tyr residues present in HSA is not accessible to the PSs. The Trp214 residue, in turn, is quenched by the ligands bound to two classes of binding sites. Possibly, these are Sudlow's site I and the heme site as they are closest to the Trp214 residue (Kamal & Behere, 2005;Sudlow et al.,1975). In each case, the quenching decreased with increasing temperature, which ensures the static type of quenching, resulting from complex formation ( Figure 5).
The Stern-Volmer constants for the quenching of HSA after the excitation at 295 nm were determined as the slopes of the curves fitted to the Stern-Volmer plots ( Figure 5, right panel). The obtained values of K SV and corresponding k q are summarized in Table 2.
Since all calculated k q values are of the order of 10 12 M À1 s À1 , and, additionally, their decrease with increasing temperature is observed, we suggest that in the investigated systems the quenching of Trp214 was not primarily caused by dynamic collision, but resulted from the formation of non-fluorescent ground-state complexes (Ouameur et al., 2005). The quenching was in each case weaker for Chlide than for Zn-Pheide, which suggests the stronger binding of the latter. However, we did not use the fluorescence quenching data to determine the exact binding constants, because due to the fluorescence enhancement at low concentrations of the PSs, these results could be inaccurate.
To investigate the conformational changes of HSA induced by the PSs, we performed synchronous fluorescence measurements. Synchronous fluorescence spectroscopy is a useful method enabling the characteristics of polarity changes in the microenvironment of the chromophores (Hosainzadeh et al., 2012;Jahanban-Esfahlan et al., 2015). Synchronous spectra are collected through the simultaneous scanning of excitation and emission at a constant wavelength interval (Dk), which equals to 15 nm for Tyr and 60 nm for Trp residues (Jahanban-Esfahlan et al., 2015, 2017. Figure 6 presents the impact of the PSs on the synchronous fluorescence spectra of HSA at Dk values of 15 and 60 nm. At low concentration (5 mM), Zn-Pheide quenched Tyr fluorescence visibly stronger and Trp fluorescence slightly stronger than Chlide. At higher concentrations, both PSs similarly reduced the fluorescence signal. No shift in the maximum emission wavelength was observed in the case of Trp ( Figure  6(C,D)), while slight blue shift (2 nm) was observed in the emission of Tyr residues in the presence of Zn-Pheide ( Figure  6(B)), indicating a decrease in polarity and increase in hydrophobicity near these chromophores (Barakat & Patra 2013;Hosainzadeh et al., 2012;Jahanban-Esfahlan et al., 2006). Interestingly, no such shift was observed for Chlide. Therefore, binding of Zn-Pheide but not Chlide leads to slight conformational changes in HSA.
As probable binding sites for the PSs, we targeted the Sudlow's sites I and II, being the major binding sites of HSA, and the heme site, chosen due to the structural similarity between the PSs and heme. The Sudlow's sites I and II are localized in the subdomains IIA and IIIA, respectively, while the heme site is in the subdomain IB (Nicoletti et al., 2008;Sudlow et al., 1975Sudlow et al., , 1976. In order to determine the binding sites of the PSs in HSA structure and compare the affinities of Chlide and Zn-Pheide, a displacement study was performed, using site-specific markers: hemin, warfarin and ibuprofen for subdomains IB, IIA and IIIA, respectively (Ascenzi et al., 2005;Baroni et al., 2001;Po or et al., 2013). Since the fluorescence intensity strongly differs between free and bound forms the PSs (Figure 3), and, at the same time, the applied site markers do not give any fluorescent signal in the range of PS emission (600 À 800 nm), molecular displacement of the PSs from HSA can be investigated by monitoring a decrease in their emission intensity. As can be seen from Figure 7, fluorescent signals of both PSs were reduced in the presence of all the site markers, though relatively high concentrations of the markers were required to significantly decrease the emission of the PSs, particularly Zn-Pheide. The most prominent displacement was observed in the case of hemin, which indicates that heme site is probably the main binding site of the investigated compounds. However, taking into account that hemin affinity to HSA is of nanomolar order, at 10 lM concentration of hemin almost all its binding sites would be occupied (Zunszain et al., 2003). At the same time, the fluorescence of the PSs still significantly exceeded that of their free forms. This confirms the assumption that several binding sites are involved in the binding of the PSs. Indeed, the emission of the PSs decreased also in the presence of two other site markers, though their relatively high concentrations were required to obtain a significant decrease. At low concentrations of warfarin and ibuprofen ( 2 lM) the fluorescence of Zn-Pheide was not reduced. Interestingly, low concentrations of ibuprofen even increased the emission of Zn-Pheide (Figure 7(C)). This effect could be explained by the fact that ibuprofen induces significant structural alteration in the heme binding cavity (subdomain IB), which may increase the affinity of Zn-Pheide to this site, similarly as it is in the case of heme (Nicoletti et al., 2008). Importantly, the decrease of the fluorescence signal was in each case higher for Chlide than for Zn-Pheide, confirming that the binding of the latter is stronger. The largest difference between the PSs were observed in their displacement by warfarin, suggesting that they considerably differ in the affinities of binding to Sudlow's site I.
Molecular docking studies were performed using the AutoDock 4.2 software to structurally investigate possible interactions sites between the PSs and HSA. Similarly to the displacement study, three probable binding sites were chosen, based on literature reports in the field of binding porphyrin-like compounds to albumin (Chaves et al., 2015;Sułkowski et al., 2016Sułkowski et al., , 2020. These were Sudlow's sites I and II, as well as the heme site. The study revealed favorable interactions (DG < 0) of both PSs with all binding sites. The cluster with the lowest binding energy was in each case the most numerous cluster. The obtained docking poses with the lowest binding energies are shown in Figure 8, and the interactions of the PSs with the residues within particular binding sites are presented in Figures 9-11.
In Figures 9-11 we show the interactions of Chlide and Zn-Pheide with the heme site and the Sudlow's sites corresponding to the lowest binding energy positions. Additionally, in Table 3 we summarize the main residues involved in the interactions with particular binding sites together with the calculated binding energies. Despite the fact that the lowest energy positions obtained for Chlide and Zn-Pheide largely overlapped (Figure 8), their interactions with the amino acid residues forming each binding site were not the same. Generally, more residues were involved in the interactions with Zn-Pheide than with Chlide, which resulted in lower values of the binding energies obtained for the former in each binding site (Table 3). The only residue interacting with Chlide but not with Zn-Pheide was Lys519, which participated in the H-bond formation in the heme site. Importantly, we did not observe the characteristic axial coordination of His, Met, Lys or Cys residues, which occurs between chlorophylls and proteins of light harvesting complexes or between heme and apoproteins of hemoglobin, myoglobin or cytochromes (Buchler et al., 1976;Hoober et al., 2007;Lu et al., 2001). Instead, binding of the metalsubstituted pheophorbides to HSA seems to be driven predominantly by hydrophobic interactions.
The PSs were found to be buried deeply inside the hydrophobic pockets of the heme site and Sudlow's site I and their binding was stabilized predominantly by alkyl-alkyl p-alkyl and p-r interactions. The largest number of hydrophobic interactions was observed for both PSs in the heme site. Amino acid residues involved in these interactions were Leu115, Arg117, Tyr138, Ile142, His146, Phe149, Phe157, Tyr161, Leu185, Arg186, Lys190 for both PSs, and additionally Arg145, Leu154, Leu182 for Zn-Pheide. Compared to the heme site, the number of residues participating in hydrophobic interactions was significantly lower in Sudlow's site I. The residues common for both PSs were in this case Tyr452, Val 455, Ala191 and Lys 195, and for Zn-Pheide additionally Trp214. Direct interaction of Zn-Pheide with Trp214 was reflected in stronger HSA fluorescence quenching by Zn-Pheide than by Chlide ( Figure 5). Only one residue involved in hydrophobic interactions in Sudlow's site II common for both PSs was identified. This was Leu 394. Three additional residues participated in these interactions in the case of Zn-Pheide. These were Leu398, Ala406 and Lys545.
In addition, at the hydrophilic entrances to the binding cavities, hydrogen bonds were formed with the participation of the PS side chains. The residues involved in these interactions were common to both PSs in the Sudlow's sites but differed in the heme site, where only one residue (Lys519) formed a H-bond with Chlide but 3 residues (Arg114, Leu115 and Glu425) with Zn-Pheide (Figure 9).
In summary, the results of molecular docking study indicate that the primary binding site for the investigated PSs is the heme site, however the binding of lower affinity is likely to occur in the other sites. Additionally, the binding of Zn-Pheide to HSA is stronger than that of Chlide, showing that a slight structural difference in the form of central metal ion exchange regulates the affinity of chlorophyll derivatives to HSA.

Conclusions
The results obtained in the present study indicate that the heme site, and to a lesser extent Sudlow's sites I and II of HSA are involved in the binding of metal-substituted pheophorbides. The binding to all identified binding sites occurs primarily via hydrophobic interactions. The determined binding affinities of both PSs to HSA are much lower than that of heme ($10 8 M À1 ) but comparable to that of protoporphyrin IX ($10 5 M À1 ), and stronger than that of other plant-derived porphyrin, pheophytin ($10 4 M À1 ), at least in one of the binding sites (Chaves et al., 2015;Kamal & Behere, (2005); Sułkowski et al., 2016). Since the identified binding sites are of the same order as that for protoporphyrin IX, it is likely that the investigated PSs may compete with this endogenous PS for binding to HSA. On the other hand, ligands with strong affinities to the Sudlow's sites and the heme site are likely to displace the bound PSs, thus increasing the concentration of their free forms. This can have a profound influence on the bioavailability of the PSs and as a consequence, on their efficacy in vivo (Szafraniec & Fiedor, 2021). Additionally, the concentrations of free PSs are likely to increase in patients with hypoalbuminemia, which is a state often associated with cancer (Navalkele et al., 2016;Takaaki et al., 2020). The affinities of binding determined for Zn-Pheide are higher than the corresponding affinities for Chlide, indicating that the central metal noticeably influences the binding. This may explain the previously observed differences in the pharmacokinetics of the investigated compounds and confirm that, similarly to animals, they are also likely to occur in humans (Szafraniec & Fiedor, 2021;Szczygieł et al., 2008). Table 3. Characteristics of the PS-binding sites in the structure of HSA predicted on the basis of molecular docking.