**3. Discussion**

The data in this study clearly shows that FGF-2 does not permeate into the multilayer but can be embedded irreversibly at a desired location within the film. Both observations are likely due to the specific interactions between FGF-2 and heparin/or heparin-like mimics, i.e., fucoidan [33,61]. The FGF-2 protein has heparin-specific binding regions on the protein surface that contain lysine residues [62]. Under physiological conditions lysine residues are positively charged and can act as hydrogen bond donors [63]. The FGF family of growth factors all have di fferent heparan sulfate glycosaminoglycan (HSGAG) binding domains [61]. The binding of FGF to HSGAGs is vital for binding to the tyrosine kinase receptors (FGFR) on cell surfaces as a ternary complex and regulating signalling [61]. FGF binding to an HSGAG oligosaccharide has been shown to involve both ionic and van der Waals forces, and was optimal due to the conformational changes of the HSGAG backbone that occur from protein binding, where the HSGAG kinks across 3 specific monosaccharide units, glucosamine-iduronate-glucosamine [61]. More recent work has shown that that binding of FGF to HSGAGs requires 3- *O*-sulfated glucosamine saccharide units, not specifically iduronate which has an additional carboxylate group in the 6- *O*-position [64]. *Fucus vesiculosus* fucoidan is predominantly comprised of long chains of 2- *O*-sulfated glucosamine units with some 3- *O*-sulfated glucosamines [50]. Thus, it is likely able to form the kinked structure around FGF proteins in a similar manner to the HSGAGs studied by Raman et al. [61]. The specificity of heparin binding domains in proteins such as FGF-2 result in quite di fferent behaviour when compared to other proteins of similar size and charge that lack the specific heparin binding domains, i.e., lysozyme [65]. Another factor to consider is the oligermisation states of FGF-2, Kwan et al. reported that HSGAGs induced dimerisation of FGF-2 via surface-exposed cystine residues [60]. The HSGAGs stabilise the FGF-2 dimers and the dimers have a more potent e ffect than the monomeric form. If the FGF-2 is dimerising when in contact with fucoidan at the surface of the multilayer, this may contribute to the lack of di ffusion into the lower layers of the film.

The work by Masuoka et al. showed that chitosan was able to protect FGF-2 from heat and enzymatic degradation at pH 7.3 (in PBS) but had no protective e ffect against acid degradation (pH < 5) [52]. This suggests that chitosan at or above its isoelectric point may be able to bind to FGF-2 as well. The amine groups of chitosan have an isoelectric point of 6.5 [66,67]. So, in PBS solution (pH 7.3) almost 50% of the amine groups will lose their positive charge. In our multilayers, if FGF-2 were binding to chitosan then when the pH is reduced to pH 5 upon return to the background electrolyte some loss of FGF-2 may be expected. However, this does not appear to be the case.

Other authors have seen similar results with other multilayer systems [68] where LYZ is able to permeate but FGF-2 does not, but, this is not always the case, and it is dependent on a multilayer structure [8,69–71]. Hsu et al. have published two works of interest, where either LYZ or FGF-2 were incorporated into the PEM structure as a component in a tetralayer [69]. In the first study, CS/poly(β-L-malic acid) (PMLA)/CS or LYZ/PMLA tetralayers were investigated with varying degrees of click crosslinking introduced via modified PMLA components to minimise interlayer di ffusion and thus, control release of therapeutics [69]. Here, LYZ was used as a model protein, and was trapped in the lower part of the PEM by a cross-linked layer of PMLA. This barrier layer was able to suppress the burst release of protein and made the release duration longer, going from 2 to 3 days. Hsu's second work, utilised the same PEM system with LYZ and FGF-2 [69]. In this study, they replaced LYZ with FGF-2 in the tetralayer PEM and found that more than six times less FGF-2 was incorporated than in similar LYZ films [69]. In addition, the loading of FGF-2 was linear with respect to film thickness. The release of FGF-2 had a similar profile to that of LYZ from the same films however, was of longer duration. The FGF-2 released from the films was found to have a greater proliferative activity than 'as-received' FGF-2, likely due to the co-release of chitosan, which may o ffer protective e ffects against heat denaturization [69].

Another group of authors who have created a body of work on the topic is Macdonald et al. who investigated PEMs comprised of a variety of synthetic and natural polyelectrolytes and their interactions with LYZ [70], FGF-2 [8] and BMP-2 [71]. In their 2008 paper, lysozyme was utilised as the polyanion in a tetralayer structure (polyX/polyanion/LYZ/polyanion)*n* where, *n* = 10–80 and polyX was one of two synthesised cationic poly(β-aminoesters), whilst the polyanion was, either heparin (HEP) or chondroitin [70]. The amount of LYZ incorporated was linear with film thickness. The same films were investigated with FGF-2 as the embedded protein in the tetralayer structure. It was found that poly2/HEP films contained the most FGF-2, partly due to the hydrophobicity of poly2 vs. poly1 which would result in less intrinsic charge compensation (poly2 films are thicker than poly1 due to this). In addition, films containing HEP can sequester more FGF-2 than the equivalent films containing chondroitin. It was proposed that the specific interactions between HEP and FGF-2 may be the dominant factor, however, in their previous work the HEP films were also able to load more LYZ than the chondroitin films [70]. So, the specific interactions between FGF-2 and HEP may not be the only reason why HEP multilayers were able to load more FGF-2 than the chondroitin films.

These studies correlate with our data confirming that heparin-specific binding sites play an important role in the uptake of FGF-2 to fucoidan/chitosan multilayer films and that fucoidan is likely acting as a heparin-mimic even when it is within a multilayer structure.

This is further confirmed by the FUC7 + PBS spectra in Figure 5, which shows that FGF-2 remains bound within the multilayer upon fucoidan adsorption, however, it is less clear in the subsequent spectra due to the overlapping amide bands of chitosan and FGF-2. Yet, it is still possible to assume the FGF-2 remains bound since upon addition of the seventh chitosan layer the amide I band does not shift as low as previously seen in earlier studies from our group [54] (to 1630 cm<sup>−</sup>1) and when the eighth fucoidan layer is adsorbed (and some chitosan is stripped) there is a shift in the amide I to higher wavenumbers i.e., closer to the peak maximum of FGF-2.

The sharp peak (Figure 5) assigned to the symmetric stretching of phosphate in PBS [58] (see electronic Supplementary Materials Figure S1) could sugges<sup>t</sup> that some phosphate remains bound via ionic interactions to chitosan after the electrolyte is changed back to KCl. Laucirica et al. have shown that amine-phosphate interactions are specific and that this binding becomes apparent under physiologically relevant conditions [72]. In addition, the same work showed that the divalent HPO4<sup>2</sup>− has an affinity for amino-groups that is five times greater than the monovalent H2PO4<sup>1</sup>− ion due to the hydrogen bonding between the protons on the amine and the charged oxygen species of phosphate ions [72]. Peng et al. found similar results with molecular dynamics that showed that phosphate ions adsorb on to amino-terminated self-assembled monolayers but chloride ions do not [73].
