Unforeseen crystal forms of the natural osmolyte floridoside

Initial identification of new polymorphic and hydrated crystal forms

Upon receiving the biomass, initial attention was drawn by a crystalline material that was easily observable on the untreated (other than initial dehydration after collection) macroalgal surface (referred to herein as untreated crystalline material, UCM, Fig. 2a, b). Remarkably, the 1H nuclear magnetic resonance (NMR) spectrum of the UCM in D2O (Supplementary Fig. 1a, b) was virtually identical to that of pure floridoside reported in the literature, giving a characteristic doublet at ca. 5.09 ppm5. This can be attributed to the equatorial anomeric proton on account of its small coupling constant (3.7 Hz), which in turn is owing to the gauche relationship between it and the axial proton on the vicinal carbon as outlined by Karplus34. Further investigation by STA (N2, 5 K min−1, Fig. 2c) identified a single concerted endothermic mass loss event (of 5.9 wt.%) at ~60–95 °C, indicating the vaporisation of a fugacious species (assumed to be water), which was then followed by a second, sharper endotherm at ca. 140 °C before thermal decomposition of the remaining organic material beginning at ~205 °C.

Fig. 2: Microscopic and thermogravimetric characterisation of untreated crystalline material.

a On the surface of the as-received Palmaria palmata surface, b following mechanical removal. Traces of the corresponding simultaneous thermal analysis (10.0 mg) c and thermogravimetric analysis (51.2 mg) d.

Separate application of higher temperature TGA (air, 5 K min−1, Fig. 2d) also highlighted the presence of a comparatively thermally stable species in the UCM which was not driven off until ~750 °C. The increase in mass over 100 wt.% is considered to be an experimental artefact related to buoyancy, as evidenced by the result of a blank run (Supplementary Fig. 2), which showed a similar pattern. Given that the second thermal event occurred well above the expected combustion temperature range of the organic content (note the ca. 9 wt.% difference at 625 °C in N2/air) and the lack of signal detected via NMR, it was concluded that it corresponded to the decomposition of a solid inorganic substance devoid of protons. It should be noted that minor amounts of other inorganics were also inferred given the small mass that remained following heating to 1300 °C (~1.5 wt.% when accounting for buoyancy). Using this information and; (i) assuming the volatile species was water and (ii) attributing the mass loss prior to 750 °C (ca. 92.5 wt.%) to either floridoside or water, a simple mass balance indicated a close to 1:1 stoichiometry of the two components (6.1 wt.% calculated).

Thermal treatment of UCM under vacuum (80 °C, ~20 mbar, 3 h) afforded a material with a similar appearance (UCM-T) and identical 1H NMR spectrum (Supplementary Fig. 1c, d) but which when subjected to same investigative techniques, yielded somewhat disparate results. No initial mass loss could be observed when UCM-T was subjected to STA however, the second endotherm was still present (albeit at ca. 136 °C) as was the decomposition from 205 °C onwards (Fig. 3a). The initial mass loss was also absent during high-temperature TGA although both decompositions remained consistent (Fig. 3b), suggesting that heating in vacuo had removed the water but otherwise unaffected the carbohydrate and inorganic components.

Fig. 3: Thermal analyses of dried untreated crystalline material and comparative p-XRD.

a Simultaneous thermal analysis (10.6 mg), b thermogravimetric analysis (50.7 mg) of untreated crystalline material dried in vacuo, c experimental p-XRD traces of selected samples compared to FI (simulated using mercury where full width at half-maximum is 0.20).

In addition to the aforementioned thermal analyses, p-XRD was also conducted on UCM and UCM-T to allow comparison with the simulated powder trace belonging to the only previously reported crystal form, FI (based on the structure deposited by Vonthron-Sénécheau et al.)11. Remarkably, both UCM and UCM-T exhibited unique powder patterns (Fig. 3c), which were clearly different to that simulated for FI (e.g., the absence of the strong signals at 2Ɵ = 7.4/19.6°) and proof that they were primarily constituted of two unreported forms of crystalline floridoside; a hydrated variant (Fh) in UCM and an anhydrous form (FII) in UMC-T. p-XRD also indicated the presence of KCl and possibly also NaCl, as can be seen by the consistent peaks at ca. 28.5/40.4/50.1° (KCl) and 31.5° (NaCl), respectively. This is in-keeping with prior reports that both species (particularly, the former) have been found to be the most abundant inorganic constituents of P. palmata23.

DSC studies on both UCM and UCM-T (Fig. 4a, b) illustrated further differences between the two materials. In the case of the former, the first of the two endotherms occurred at a similar temperature to that recorded using STA, which was followed immediately by an apparent exothermic deviation from the baseline at ca. 87.5 °C and then a second, significantly broader endotherm from 93 to 118 °C. During the second heating cycle, there is a clearly observable glass transition at −12.4 °C, signifying the presence of an amorphous melt (which then appeared to undergo partial cold crystallisation and subsequent re-melting upon heating). Conversely, UCM-T exhibited only a single broad endotherm at ca. 115–142 °C and a well-defined glass transition (Tg) of 30.1 °C, highlighting the considerable plasticisation by water which is comparable to that reported for other molten carbohydrate matrices35.

Fig. 4: Differential scanning calorimetry of Palmaria palmata-derived materials.

First heating, first cooling and second heating cycles of; a untreated crystalline material (10.7 mg), b untreated crystalline material dried in vacuo (10.0 mg), c untreated crystalline material heated to 90 °C before cooling (11.8 mg), d simultaneous thermal analysis of the latter (11.0 mg).

In a further experiment, UCM was heated to a temperature of 90 °C (i.e., past the initial endotherm) within a sealed DSC pan prior to cooling (UCM-90, Fig. 4c) before being removed and subjected to STA (Fig. 4d). In this case, it was difficult to observe any obvious glass transition (at least one above −80 °C) upon cooling, indicating a comparative lack of amorphous matter. Subsequent STA of UCM-90 highlighted the presence of residual water which unlike UCM, was lost gradually and over a much broader temperature range starting at ca. 30 °C. In addition to this, the only notable endotherm (aside from the omnipresent decomposition signal at >205 °C) in the spectrum was similar (albeit broader in part due to residual water vapourisation) to that detected in UCM-T, indicating the presence of FII. This supposition was corroborated by attenuated total reflection Fourier-transform infrared (ATR-FTIR) analyses (Fig. 5) of UCM, UCM-T, and UCM-90 wherein it can be seen that the spectrum of thermally untreated UCM is markedly different to that of both heated samples, which instead are essentially identical.

Fig. 5: ATR-FTIR spectra of Palmaria palmata derived materials.

a Untreated crystalline material, b untreated crystalline material dried in vacuo, c untreated crystalline material heated to 90 °C within a sealed DSC pan. Insets magnify the 1550–1750 cm−1 region for a and c. Arrow indicates increasing transmittance.

Interestingly, the only discrepancy between the spectra of UCM-T and UCM-90 is the presence of a somewhat broad but weak vibrational band at  ≈ 1645 cm−1, which is attributed to the deformation modes of H2O, consistent with residual water. The same absorption band is also observable in UCM, wherein it manifests at a higher wavenumber (1672 cm−1) and takes a considerably sharper form, indicating a clear difference in the water environment. This value is in-keeping with those of water deformation in similar carbohydrate crystals such as trehalose dihydrate (1678/1639 cm−1) and maltose monohydrate (1639 cm−1)36,37. This is further indication that the series of DSC signals at <93 °C in UCM correspond to the deformation of Fh and eventual formation of crystalline FII (then liquified upon further heating cf. broad endotherm) although the exact sequence and nature of the thermal events is not yet fully elucidated and should be investigated further.

Physicochemical characterisation and future outlook

On account of the high purity of UCM, it was possible to generate all of the floridoside polymorphs using basic heating and solvent manipulations. The typical approach consisted of boiling UCM in MeOH (initial concentrations of ca. 25 mg mL−1) and hot filtration of the resulting suspensions followed by either storage or further treatment of the filtrate. The most reliable method of generating FII was via prolonged boiling (for several minutes) of UCM before hot filtration and natural cooling in sealed vials for 1–7 days. This generally led to clear, colourless crystals, which either grew as well-formed cuboid-like (Fig. 6a) or higher (Fig. 6b) polyhedrons, presumably depending on the magnitude of the driving force toward crystallisation.

Fig. 6: Isothermal optical micrographs of crystalline floridoside.

a and b FII, c, and d FI, e Fh.

Conversely, immediate addition of an excess of EtOAc (as an antisolvent) to the methanolic filtrate resulted in the formation of FI, either concomitantly with FII or as the sole crystalline form depending on the relative quantities of the EtOAc and filtrate. In the former case, diffractable needle-like crystals (Fig. 6c) could only be prepared via addition of relatively smaller amounts of EtOAc (~3:1, v/v), which resulted in the formation of clear solutions following antisolvent addition and from which, the FI/FII crystals could then be manually separated. Increased addition of EtOAc so as to induce turbidity (~5:1, v/v) reliably led to the formation of spherulite-type clusters containing very fine needles (Fig. 6d), which were collected quickly following natural settling of the suspension (2–3 h) in order to minimise any potential solvent-mediated polymorphic transformation to FII. It should be noted that further increasing the amount of EtOAc resulted in the liberation of a globular precipitate from solution, which likely corresponded to amorphous matter, although this was not investigated further.

In the case of Fh, large colourless crystals (Fig. 6e) could be obtained via partial slow evaporation (over several days) of the MeOH filtrate under a fume hood at room temperature. However, this approach was somewhat aleatoric and occasionally resulted in the generation of anhydrous forms or a liquid that was devoid of crystals. This is indicative of a delicate interplay between the relative concentrations (viz. thermodynamic activities) of the filtrate constituents (namely, floridoside, H2O, MeOH, and salts) as has been reported for the crystallisation of other water-soluble solutes from alcohol/water solutions38.

The newly identified crystal structures of FII, Fh are displayed alongside FI (for which the CCDC data provided by Vonthron-Sénécheau et al. has been used) in Fig. 7 (full data is given in Supplementary Table 1)11. All three crystals are orthorhombic and belong to the P212121 space group with each unit cell comprising four floridoside molecules in addition to four H2O molecules in the case of Fh as a stoichiometric monohydrate. The unit cell axes (Supplementary Fig. 3) of both FII (a = 8.54811(10) Å, b = 9.19251(10) Å, c = 14.34851(17) Å) and Fh (a = 8.22038(16) Å, b = 11.2533(3) Å, c = 12.9852(2) Å) display relatively comparable lengths when compared with FI, wherein one axis is markedly longer and another noticeably short (a = 4.88440(10) Å, b = 9.7259(3) Å, c = 23.8754(6) Å) on account of an almost head-to-tail arrangement of molecules, which results in a very high aspect ratio. This is likely to be the origin of the consistent needle-like morphology, which was observed for all of the FI crystals.

Fig. 7: Thermal ellipsoid representations for FI, FII, and Fh.

Ellipsoids are shown at the 50% probability level wherein hydrogens are drawn with arbitrary radii. Torsion angles ({upomega}) and ϕ are depicted to aid visualisation.

From a molecular perspective, the pyranose ring adopts the 4C1 conformation in all three forms, with each existing as gauche-trans rotamers with respect to the C5-C6 bond, displaying a O6-C6-C5-O5 torsion (({upomega})) angle of 67.99(7), 64.49(18), and 58.27(18)o in FI, FII, and Fh, respectively, (structural overlays are shown in Supplementary Fig. 4). Interestingly, this conformation (as opposed to gauche-gauche or trans-gauche) appears to be common across many naturally occurring α-substituted d-galactopyranoses (e.g., methyl α-d-galactopyranoside monohydrate = 62.60°39, α-d-galactopyranosyl-α-d-galactopyranoside = 63.83°40, α,β-melibose monohydrate = 65.69°)41. Aside from the obvious differences resulting from the additional presence of water (in Fh), the most discernible structural discrepancies of the floridoside molecule originate from the orientation of the pyranose-based hydroxyl groups and configuration of the glycerol moieties. The latter is exemplified in the case of the glycosidic bond, whereby the torsion angle (ϕ, O5-C1-O1-C7) of Fh at 91.02(16)o is considerably higher than that of either FI (73.89(6)°) or FII (63.39(17)°) and sufficiently so as to enable the rotation of a terminal glycerol OH toward O5/O6 (Fig. 7).

Satisfyingly, p-XRD patterns simulated from the single crystal data (Supplementary Fig. 5) of Fh and FII were very similar to the experimental traces recorded for UCM and UCM-T, again confirming their identity and surprising purity. This was also the case for the corresponding ATR-FTIR spectra (Supplementary Fig. 6), which were identical to those presented in Fig. 5 but both wholly different to the spectrum of FI. In addition to these spectroscopic discrepancies, the thermal properties of each of the three polymorphs were also found to be considerably disparate—as illustrated by DSC and hot-stage optical microscopy (Fig. 8). In the case of FI (Fig. 8a), a clear melting endotherm with an extrapolated onset (To) and maximum fusion temperature (Tm) of 128.1 and 130.0 °C can be observed during the first heating cycle, consistent with previous reports (127–134 °C)10,11. This is followed by a smaller exothermic signal, which is attributed to the formation of crystalline FII, which then proceeds to melt at ca. 140 °C—in good agreement with the To (139.3 °C) and Tm (142.2 °C) found for the pure FII crystal (Fig. 8b). In both cases, there is clear indication of glass formation upon melt quenching, with a reproducible Tg of ca. 32 °C. The aforementioned results also seem to indicate a monotropic relationship between the FI and FII polymorphs, whereby the irreversible transition from the meta (FI) to stable (FII) dimorph occurs either through a solid–solid or liquid–solid process close to or above the fusion T of the former. It should be noted that neither such transformation was observable during hot-stage microscopy of FI (a full sequence of micrographs from 45 to 139 °C is provided in Supplementary Fig. 7) however, which is not unexpected given the kinetic nature of these events. In both crystals systems, it was also possible to observe melting of a comparatively small quantity of FI at ca. 130 °C during the second heat step (see insets), which solidified from the amorphous phase during the heat/cool/heat cycle.

Fig. 8: Cyclic DSC and hot-stage micrographs of floridoside crystals.

a FI (8.2 mg), b FII (8.8 mg), c Fh (6.0 mg). Insets within the DSC traces highlight the 100–150 °C a, b or 75–130 °C c regions wherein y axis ticks correspond to 0.5 mW.

Interestingly, the large, well-formed Fh crystals exhibited only a single large endotherm with To and Tm values of 86.0 and 89.1 °C (Fig. 8c) upon heating to 150 °C. Heating past this signal (to 100 °C) followed by immediate cooling led to the formation of a glass with a Tg of ~−11.2 °C (which underwent cold crystallisation during the 2nd heat cycle—see Supplementary Fig. 8), similar to that found for molten UCM (−12.4 °C), confirming that a significant solid–liquid transition had occurred. Indeed, this was also corroborated by the observation of an isotropic melt during hot-stage optical microscopy, which subsequently persisted through further heating (the entire experimental range of 35–135 °C is shown in Supplementary Fig. 9). This thermal behaviour is notably different to that described earlier for UCM (also constituted of Fh), wherein crystalline FII was instead formed upon initial heating. It is tentatively hypothesised that this discrepancy is related to the greater structural integrity of the well-formed, single Fh crystals, which ultimately results in comparatively retarded molecular dynamics of the crystalline components upon deformation, preventing expedient lattice rearrangement, which is necessary for the formation of the anhydrate. It is likely that such behaviour is highly dependent on experimental conditions, as has been reported for the thermally induced deformation of trehalose dihydrate and which seemingly necessitates the need for more detailed future studies42.

Unsurprisingly, given the abundance of OH groups in the floridoside molecule, all three crystals exhibit a complex spatial network of hydrogen bonding (H-bonding). To probe for potential differences in the H-bonding of the three different forms, a relatively lenient (albeit arbitrary) criteria was applied, wherein the maximum distance and minimum angle between the OH donor hydrogen and acceptor (R1-O-R2) were limited to ≤3 Å and ≥90°, respectively. Extending the definition past these limits led to a rapid increase in the number of possible contacts, many of which appeared to be unrealistic and hence, were not considered as was also the case for C-H donors (which have been reported to exist in similar systems)43. Interestingly, these values have some basis in previous relevant literature with Steiner and Saenger independently concluding that it was difficult to discriminate between H-bonding and non-bonding regions outside the same cutoffs during their analysis of 15 different non-ionic carbohydrates44. Indeed, the geometric boundaries currently employed may actually be overly inclusive, as highlighted by the fact that the H2O molecule in Fh unrealistically forms five H-bonds (Supplementary Fig. 10).

Enacting the aforementioned criteria, the H-bonds were found to fall into one of two general groups. The first would conventionally be described as ‘strong’, displaying lengths and angles of ≤2.20 Å and ≥140°, respectively. The second, which are comparatively ‘weaker’ in nature are >2.20 Å and <140° and likely to be more ambiguous with respect to their actual manifestation within the crystal. It is interesting to note that all hydroxyls in FI are involved in both strong intermolecular hydrogen bond donation and acceptance (Fig. 9d), as highlighted previously by Vonthron-Sénécheau et al.11 but also that only a single weak intramolecular H-bond could be identified when using the same description (Fig. 9a). In contrast to FI, FII was found to display greater and considerably more varied H-bonding, containing many bi-/trifurcated configurations (both intra/intermolecular), which may contribute to the greater stability of the latter as reflected in the higher fusion temperature. It should be noted however, that the sharp peak at ca. 3576 cm−1 in the IR spectra (Fig. 5 and Supplementary Fig. 6) indicates the presence of a single, isolated hydroxyl group, which again suggests that not all of the H-bonds identified in Fig. 9 actually manifest within the crystal.

Fig. 9: Calculated H-bonding profiles for floridoside crystals.

Showing ‘strong’ (<2.20 Å and >140°) and ‘weak’ (2.20–3.00 Å and 90–140°) intramolecular contacts (strong = dark green dashed line, weak = light green dashed line) found for a FI, b FII, and c Fhand also intermolecular contacts (strong = navy dashed line, weak = cyan dashed line) found for d FI, e FII, and f Fh. Graphically summarised in g (intramolecular H-bonds) and h (intermolecular H-bonds). Calculations were made using Mercury wherein only OH donors/O acceptors and intramolecular contacts between atoms separated by >1 bond were permitted.

In Fh, all but one of the OH groups are engaged in strong intermolecular donation and acceptance with either a second floridoside or water molecule, with the only exception being that one of the glycerol hydroxyls instead acts as a strong intramolecular donor to O6, around which sufficient space exists owing to the aforementioned wider glycosidic torsion angle (Fig. 9c). The same hydroxyl forms a second, weaker contact to the ring oxygen with two further instances of weak intramolecular H-bonding (O(3)H···O(4) and O(2)H···O(1)) also being detectable in Fh.

It is hypothesised that the formation of this hydrate may have some utility in vivo as a means of regulating the intracellular concentration and distribution of water, akin to what has been proposed for trehalose dihydrate25. This would provide an effective control of internal osmotic pressure, especially within specific domains wherein the local floridoside content may be particularly high, such as the cytoplasm45,46. In addition to this, some authors have also implicated floridoside synthesis as a mitigation strategy for desiccation stress (during low tide) in various red macroalgae, although this may not be universally adopted across all species4,47,48,49. This again is very reminiscent of the trehalose accumulation that has been reported in other organisms (tardigrades, nematodes etc.) during periods of anhydrobiosis. It has been postulated that the functionality of trehalose could be related to the reversible un/loading of water within structured channels that exist within the dihydrate25,31. Inspection of the Fh lattice indicates that the formation of a connective water network is unlikely given the large distance between neighbouring H2O molecules (≥6.56 Å O···O), suggesting a fundamental difference in the protective mechanism at the molecular level. This is further supported when considering that the glass transition of floridoside is far lower than that reported for trehalose (ca. 32 °C vs. ca. 100–115 °C), which makes in vivo vitrification by the former far less realistic in comparison50.

Concerning anthropocentric use of floridoside, the identification and characterisation of metastable and hydrated crystalline species is noteworthy given that individual polymorphs and hydrates often exhibit very disparate physicochemical behaviours as exemplified by other carbohydrates such as d-glucose and d-lactose51,52. The metastable crystal of a dimorphic pair often displays a comparatively faster dissolution rate but greater hygroscopicity for instance, which suggests that controlled crystal engineering towards either FI, FII, and/or Fh should facilitate more optimised floridoside-containing product/processes that can be tailored towards an individual end application. Given this, more detailed studies that further explore the influence of crystalline structure on such properties will be invaluable and hence, have been initiated.

In conclusion, despite its extensiveness in nature and significant history, there has remained a lack of detailed characterisation concerning the preeminent compatible solute floridoside. Herein, we have presented unambiguous characterisation of two unforeseen crystal forms of this important biomolecule; an unheralded anhydrous polymorph and also a monohydrated variant, which has seemingly remained elusive until now. The acknowledgement of these newly identified forms of floridoside—which display distinct physical and thermal properties, is of considerable industrial and academic significance given both the increasing commercial interest and recognition of its role in vivo as a promising versatile and natural compound.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *