Graphene oxide modulates inter-particle interactions in 3D printable soft nanocomposite hydrogels restoring magnetic hyperthermia responses.

Hydrogels loaded with magnetic iron oxide nanoparticles that can be patterned and which controllably induce hyperthermic responses on AC-field stimulation are of interest as functional components of next-generation biomaterials. Formation of nanocomposite hydrogels is known to eliminate any Brownian contribution to hyperthermic response (reducing stimulated heating) while the Néel contribution can also be suppressed by inter-particle dipolar interactions arising from aggregation induced before or during gelation. We describe the ability of graphene oxide (GO) flakes to restore the hyperthermic efficiency of soft printable hydrogels formed using Pluronics F127 and PEGylated magnetic nanoflowers. Here, by varying the amount of GO in mixed nanocomposite suspensions and gels, we demonstrate GO-content dependent recovery of hyperthemic response in gels. This is due to progressively reduced inter-nanoflower interactions mediated by GO, which largely restore the dispersed-state Néel contribution to heating. We suggest that preferential association of GO with the hydrophobic F127 blocks increases the preponderance of cohesive interactions between the hydrophilic blocks and the PEGylated nanoflowers, promoting dispersion of the latter. Finally we demonstrate extrusion-based 3D printing with excellent print fidelity of the magnetically-responsive nanocomposites, for which the inclusion of GO provides significant improvement in the spatially-localized open-coil heating response, rendering the prints viable components for future cell stimulation and delivery applications.


Introduction
Magnetic nanoparticles (MNPs) are currently used and are under development for multiple biomedical applications including cancer treatment, as delivery/release agents in vitro and in vivo and as contrast agents for magnetic resonance imaging, due to their controlled responses to applied magnetic fields which are strongly dependent on the particles dispersion state [1][2][3][4][5]. Graphene and graphene oxide (GO) are being widely explored for liquid crystal [6] and electronics applications [7], and as composites [8] often for tissue engineering as they can provide functional responses and can modulate the aggregation state of other nanocomposite components [9,10], providing means to control the final physical (e.g. mechanical or thermal) properties. Materials studied in this regard include hydrogels [11][12][13][14].
Hence the intrinsic functional potential and controlled component interactions arising from including, for instance, MNPs and GO flakes as nanocomposite fillers within polymer matrices offers opportunities for designing responsive systems. Mourdikoudis et al. recently reviewed synthesis, physicochemical properties, and applications of magnetic nanocomposites, noting the use of graphene and GO as support surfaces for immobilizing MNPs as an emerging route to functional nanocomposites [15].
Amongst possible matrix formers, pluronics, or poloxamers, are thermoresponsive triblock co-polymers, composed of hydrophilic poly(ethylene) oxide (PEO) blocks flanking a more hydrophobic central poly(propylene) oxide (PPO) block, commonly used for stabilizing nanomaterials and forming soft hydrogels. The variation in hydrophobicity, which is block length-dependent, renders the materials thermoresponsive [16,17]. Above a phase transition temperature well-defined micelles are observed with PPO blocks forming a hydrophobic core and PEO a hydrophilic shell [17][18][19], as a result gels are typically observed at room temperature for ≥ 21 w/v%, depending on molecular weight. Pluronics provide a tuneable 'soft' shear-thinning matrix with is usually injectable or printable at room temperature, with Pluronic F127 (or F127, of average formula PEO101-PPO56-PEO101, with the sol→gel transition at ~4 o C and excellent room-temperature shear-thinning) a particularly common choice.
GO, MNPs, and some combinations of these have been investigated as nano-fillers in F127 matrices with the focus of most studies on understanding the inter-particle interactions [20][21][22][23][24]. The addition of GO alone to F127 provided a responsive injectable hydrogel, capable of undergoing phase transition upon infrared (IR) irradiation or change of temperature or pH [21]. The encapsulated GO flakes were described as inducing changes in the polymer self-assembly due to stronger hydrophobic GO-PPO interactions, and with the hydrophilic poly(ethylene) oxide (PEO) chains extending into surrounding water [20,21].
This reflects the consensus view as recently reviewed by Alegret et al. [25]. Other related reports focusing on biomedical application include Li et al. who described MNPs both physically [22] and chemically [23] linked to GO, encapsulated within an F127 coating to form hybrid micellar nanocomposite suspensions (nanogels) as carriers for pH-dependent doxorubicin delivery [22], and as magnetic resonance imaging-trackable Dox carriers [23].
Bulk hydrogels containing MNP-functionalized GO were also developed for combined again Dox release and hyperthermic therapy [26].
AC magnetic field heating of MNPs (hyperthermia) arises due to coupling of the particle moment with the applied field due to particle rotation (Brownian contribution) or moment re-orientation (Néel contribution). The hyperthermic efficiency can be quantified using the specific absorption rate (SAR) of AC-field energy, under quasi-adiabatic conditions, and is commonly used as a comparison between samples, SAR (W g -1 ) is defined as; (1) where C is the volumetric heat capacity of the medium (J mL -1 o C -1 ), Vs the suspension volume (mL), mFe the mass of iron (g) obtained using flame atomic absorption spectroscopy and [ ] =0 ( o C s -1 ) defines an initial slope of the temperature-time plot usually extracted as the linear term from a polynomial fit, a commonly accepted approach [27], stemming from the presence of radiation and its fourth-power temperature dependence and the desire to weigh the linear part to the earliest (close to adiabatic) part of the response [28]. We used a fourth order polynomial, as is common practice [27], see Methods. Encapsulation of MNPs within hydrated networks and hydrogels suppresses any Brownian SAR contribution by constraining particle motion [29], while any Néel contribution can be reduced by interparticle dipolar interactions due to aggregation induced before or during gelation [30].
We recently described nanocompositing of MNPs, in this case unmodified magnetic iron oxide nanoflowers (NFs), with F127 to form 3D-printable hydrogels that provided spatiotemporally-controlled temperature increase and associated dye release, on AC-field irradiation [24]. We found that the SAR value was indeed suppressed on gelation, although sufficient response was observed at high NF concentration (Fe 140 mg mL -1 ) to demonstrate the principle. Our interpretation was that the Néel contribution, which is known to dominate over the Brownian for nanoflowers [31] was partially suppressed due to some aggregation.
Higher SAR values are preferable as they allow for better control over the hyperthermic response (higher temperature jump realised in the applications format) and the use of lower particle concentration which may facilitate, for instance, UV-initiated crosslinking for downstream processing.
In this study we evaluated the possibility of improved nanocomposite hydrogels comprised of F127, NFs and GO. It was anticipated that planar GO flakes would mediate F127-induced NF aggregation, maintaining NF separation, and recovering the high hyperthermic potential of NFs. SAR values were screened for nanocomposite suspensions and gels of different composition, to elucidate the role of GO depending on the NF surface chemistry (unmodified or PEGylated). It was shown to be possible to recover most of the native SAR under certain conditions. Finally spatial patterning of the magnetic F127-GO nanocomposite hydrogels was demonstrated with significant improvement in hyperthermic response, as compared to GO-free equivalents, observed when using an open coil AC-field system suitable for cell culture or other biomedical applications.

2.1.Formation of nanocomposites
Magnetic iron oxide nanoflowers (NFs) were synthesized using a minor adaptation of the procedure developed by Hugounenq et al. [32]. A typical synthesis, see Methods, provides NF suspensions with a Fe yield of ~49%. The TEM size of 29 ± 3 nm ( Figure 1A) obtained for NFs is similar to our previous report [24] and the low average hydrodynamic size of 37 nm and low polydispersity (PDI) of 0.17 obtained at pH 4 demonstrate full particle dispersion ( Figure 1A and E). Suspensions of unmodified as-produced, or bare, NFs (here named BNFs) are not stable at physiological pH of 7 due to the absence of net surface charge (Figure S1B), as is most often required for biomedical applications and nanocompositing. Therefore polyethylene glycol (PEG) chains of 8000 Da were grafted to the NF surfaces, following the protocol described by Studart et al. [33] (Figure 1B and S1A-C), see Methods, forming PEGylated NF suspensions (named PNFs). PNFs are found to be 41 ± 4 nm in size (mean and standard deviation of measurements for 30 PNFs) from the amplitude modulation mode atomic force microscopy (AFM) height image ( Figure 1C). The corresponding magnetic force microscopy (MFM) phase image recorded during lift-mode operation shows a phase shift associated with the resonance curve shift due to the long-range magnetic force gradient arising from PNFs ( Figure 1D). Surface coating was evident from the increase of hydrodynamic size from c.37 nm at pH 4 for BNFs, to c.58 nm (again with low PDI of 0. 18) for PNFs at neutral pH ( Figure 1E), which is in the expected range given the graft chain length. Steric stabilization of PNFs is also confirmed by the close to neutral zeta potential (ζp) at pH 7, and by the broad pH range, from 6-10, over which the hydrodynamic size remains unchanged ( Figure S1C). Full particle dispersion of these single-component suspensions is also evident from the hyperthermic properties, see below.   4) and sPNF (at pH 7), respectively.
Graphene oxide was synthesized using a modified Hummers' method [34], see Methods. Here, we considered small GO flakes of size 240 ± 20 nm (a full characterization of the flakes is provided in Methods and shown in Figures S2 and S3). GO flakes are typically sterically-and charge-stabilized across pH values up to basic with ζp of -42 ± 3 mV at pH 7 ( Figure S3A) [35]. BNF, or PNF, and GO suspensions were mixed at the appropriate pH to ensure dispersion (Scheme 1A) and in different ratios, see Methods, to form stable co-suspensions (named sBNF_GO or sPNF_GO, see Table 1). The sPNF_GO co-suspensions displayed homogeneity and stability at pH 7, indicating weak interactions between PNFs which at neutral pH have ζp close to zero (Figure S1B), and GO flakes which at neutral pH have ζp of -42±3 mV ( Figure S3A). Increasing the amount of GO in the co-suspensions induced an apparent ζp decrease to close to that of the single-component GO suspension (Figure 2A)  To form nanocomposite hydrogels F127 powder was dissolved in sBNF_GO or sPNF_GO (MNF-GO co-suspensions), at the appropriate concentrations, followed by gel-solgel re-fabrication by cycling from room temperature to 4 o C and back to room temperature.
The resulting gels were stable and appeared homogeneous even after a single cycle (Scheme 1B and 1C) and are named here as gBNF_GO or gPNF_GO (see Table 1). SEM-EDX revealed similar distribution of Fe signal (~0.3 at%) across all gel samples, originally prepared at a Fe concentration of ~2.7 ± 0.3 mg mL -1 ( Figure S4). These findings show that the nanocomposite gels are homogeneous as was further confirmed by hyperthermia measurements (next section). Finally, all electron micrographs show the porous structures expected for hydrogels ( Figure S4). The magnetic properties, and in particular the AC-field hyperthermia responses, of the nanocomposite suspensions and gels are described below.

NF_GO suspensions
It is generally found that for MNP and NF syntheses that use conventional heating-up or polyol approaches the measured SAR values vary significantly from batch to batch (reflecting the stochastic nature of the nucleation process, and other factors that are difficult to control) [36,37]. Furthermore, irrespective of the synthetic approach used, the 'native' SAR values (i.e. of single component NF suspensions) obtained can be concentration dependent (due to dipolar interactions, usually arising from partial aggregation) and the position of the thermal probe in the sample (due to inhomogeneity, or even partial precipitation, often associated with partial aggregation) [28]. In this study SAR was determined at AC 535 kHz, IAC 24 mT. The concentration dependencies of SAR were explicitly studied and care was taken to confirm sample homogeneity and stability of the measurements. The error given throughout for individual SAR measurements is c.5%, stemming from uncertainty in extracting the initial slope and the Fe concentration, see Methods. The SAR response of exemplar dispersions and hydrogels was also tested as a function of probe depth, it was found that all measured values fall within ±5% of the average for co-suspensions and ±2.5% for hydrogels, see Methods and Supporting Information (Table   S1), confirming the absence of sedimentation.
Variability of the native SAR is well-known (if rarely discussed) and is difficult to avoid, hence we took the approach of preparing multiple NF batches for which the values were indeed found to vary (in the range 180-320 W g -1 with median around the 220 W g -1 ) and evaluated the effect of NF and GO content on SAR for each batch. For multiple batches we observed and report consistent effects on SAR arising from varying the NF and GO content, irrespective of the native SAR. Hence the findings described are general and not related to the particular magnetic properties of any given NF batch. Note also that the heat capacity of H2O was used in all SAR calculations presented in this study; the validity of this assumption for suspensions and gels is discussed below.
For sBNF and sPNF (single-component) suspensions, the measured SAR values are independent of the Fe concentration ( Figure 3A-B). For the sBNF (bare NFs in suspension) batch shown in Figure 3A an average SAR value of 200 ± 10 W g -1 was obtained across a 40-fold NF concentration range. We also observe minimal effects of PEGylation on the SAR value for the single component suspensions, as expected for fully dispersed particles.
Similarly for different sPNF (PEGylated NFs in suspension) batches, shown in Figure 3B, a SAR value of 220 ± 4 W g -1 was obtained over a wide Fe concentration range. This also demonstrates that there is minimal effect of nanoflower concentration on the heat capacity of the suspensions, as expected given the dilution.
On the other hand for sBNF addition of GO flakes consistently led to a significant decrease of the (sBNF_GO co-suspension) SAR values, irrespective of the pH of the stable sBNF (bare NFs only suspension) used, i.e. pH 4 ( Figure 3C) or pH 12 ( Figure 3E), for which the native sBNF ζp values were +34 and -49 mV, respectively ( Figure S1B). The observation of SAR suppression, irrespective of the sign of the BNF surface charge, suggests GO-induced BNF aggregation by hydrophobic interactions; hence this is most likely due to πelectron rich sp 2 -hybridized regions on the planar GO surface interacting with γ-Fe2O3 sites inducing BNF-BNF interactions on-flake [38]. The inter-NF dipolar interactions incurred suppress the Néel contribution, which has been shown to dominate over any Brownian contribution for NFs [31]. On the other hand, we did not observe any SAR suppression for sPNF_GO (co-suspensions), irrespective of whether GO or NF concentration was varied ( Figures 3D and 3F); the SAR of these co-suspensions was independent of Fe concentration in this range. At constant GO concentration of 0.5 mg mL -1 , an average value of 213 ± 13 W g -1 ( Figure 3F) was measured, which is similar to the native SAR of this batch. Similar observations were made for three other independently prepared batches irrespective of their native SAR ( Figure S5). These observations strongly support the interpretation that hydrophobic interactions generate partial aggregation in co-suspension for sBNF_GO, but not for sPNF_GO. Note that the effect of GO, in the concentration range used, on the heat capacity of both suspensions and soft hydrogels of this type has been shown to be weak. For example, a marginal ~12%, decrease in heat capacity was shown on adding GO at 1 mg mL -1 to Pluronics-based gels [20]. A decrease in heat capacity induced by progressively increasing GO content would decrease the SAR value extracted (see Equation 1). For sPNF-GO cosuspensions, Figure 3D shows that there is no such decrease up to GO content of 1 mg mL -1 .
Note that we have also demonstrated the lack of any AC-field induced heating for pure GO suspensions ( Figure S6). In the next section, we describe consistent increases (SAR recovery) for nanocomposite hydrogels with increasing GO content, hence we can safely ascribe these changes to differences in inter-flower interactions mediated by GO flakes, not to changes in heat capacity.

NF_GO nanocomposite hydrogels
Considering the single-component gels, for a typical gBNF sample (gels with bare NFs, without GO, at pH 4), a SAR value of 50 ± 14 W g -1 was measured across the Fe concentration range (Figure 4A), which is a significant, 85%, suppression as compared to the equivalent sBNF suspension (native SAR 309 W g -1 ). Similar suppression was noted for gBNFs in pluronics in our earlier study [24]. While for the equivalent gels with PEGylated nanoflowers, gPNF, (without GO, at pH 7) SAR decreased by only ~35% from 221 ± 3 to 145 ± 4 W g -1 and was independent of Fe concentration ( Figure 4B). This indicates that    Figure   S7; (iv) 3 data points in A were reproduced from reference [24] Figure 4C), and at constant flake concentration (GO 0.5 mg mL -1 , Figure 4D). The SAR for the gPNF batch (without GO) used to generate these two plots was 122 W g -1 as indicated in Figure 4C, and the native suspension value (without GO) was 188 W g -1 . SAR was found to progressively increase with addition of GO flakes up to 182 W g -1 at GO/Fe c.0.32 (corresponding to 2.5 mg mL -1 Fe and 0.80 mg mL -1 GO), i.e. the SAR of the nanocomposite gel recovered to a value similar to that of the native GO-free suspension. The fact that it is possible to almost fully recover the native suspension SAR for gPNF_GO demonstrates again that any contribution from Brownian processes to the hyperthermic responses of NFs is negligible [31]. It also again shows that it is a reasonable approximation to use the heat capacity of H2O in calculating SAR for the nanocomposite gels.
Similar SAR measurements as a function of GO were performed on three other independent PNF batches with varying native SAR values ( Figure S7). A consistent increase in SAR with increasing GO content was observed irrespective of the native SAR, with all batches reaching a maximum (batch dependent) value at GO/Fe in the range of 0.3-0.38 (corresponding to 1.08 mg mL -1 GO). This confirms that higher GO content is invariably associated with recovery of higher SAR values, to close to that of the corresponding sPNF_GO co-suspensions (Figures 3D and S5). The effect of GO on the gels is also clearly demonstrated by comparison of sPNF_GO ( Figure 3F) in which case the SAR is independent of GO concentration with gPNF_0.5GO ( Figure 4D) in which case SAR decreased with increasing Fe concentration over the same Fe concentration range. Note cosuspensions and nanocomposite gels with GO > 1.20 mg mL -1 were difficult to prepare due to constraints on initial NF and GO concentrations in the master batches, and co-suspension stability at higher concentrations.
In summary, it is clear that GO-induced interactions with PPO segments reducing inter-PNF interactions, and hence inter-flower dipolar coupling, are a general feature of this nanocomposite gel system. Within the gels the presence of the GO flakes alters the molecular interactions and hence the arrangement of the nanocomposite components. We suggest that hydrophobic attraction between the PPO blocks in F127 and the exposed GO surface decreases the potential for non-cohesive interactions between PNFs and those PPO blocks.
GO-induced higher SAR values measured at lower Fe concentration, as compared to the GOfree case, strongly support increasingly cohesive (hydrophilic) PNF-F127 interactions, preventing inter-flower dipolar coupling which would reduce hyperthermic efficiency [30].
To our knowledge there is only one related report of the effect of MNPs and GO on hyperthermic responses in gels. Le et al. [39] investigated magnetic manganese ferrite MNPs immobilized onto GO. This system is complicated as MNP binding was described as increasing the magnetic anisotropy (increasing SAR), while gelation was suggested to decrease the Brownian contribution which, unlike NFs, is significant for spherical particles in this size range. Differences in the concentration-dependencies of SAR were also described

Spatiotemporally specified AC-field induced heating in patterned PNF_GO nanocomposite hydrogels
Finally, following our initial report of AC-field induced heating in structures printed using gBNFs as the magnetically-responsive ink [24], a preliminary evaluation of the printability of gPNF and gPNF_GO and the responses of the prints was undertaken. In all cases the nanocomposite gels were found to be shear-thinning and printable and, as expected, the presence of GO improved the AC-field response. It was observed (Figure 5, Figure S8) that samples with and without GO provided smooth prints using a standard extrusion format without significant optimization, and the deposited material showed good structural integrity ( Figure 5)  In the case of gPNF (without GO) the hyperthermic response (evaluated at higher frequency, AC 663 kHz, and lower field strength, IAC 16 mT, than for the previous measurements, see Methods) was weak, with an increase in temperature after 600 s, T600s, of ~0.4 °C observed (Figure 5A), as we have previously observed [24]. Upon addition of GO a significant increase in response was observed, with T600s of ~2.5 °C at the plateau for gPNF_0.6GO ( Figure 5B) and gPNF_1.08GO (Figure 5D). A weaker response was observed for gPNF_0.8GO (Figure 5C), even though its native SAR was slightly higher, due to greater dissipation associated with smaller FT and hence higher surface area (Figure 5 and Figure S9). Interestingly, for the three GO-containing prints shown in Figure 5, the extracted initial slope (the average for the whole grid area, blue curves in Figure 5), or T/t0, is independent of GO content, with an average value of 0.012 ± 0.001 °C s -1 obtained. Note that this is for samples of similar native SAR and the same Fe concentration. We have not scaled T/t0 by Fe content to generate a 'SAR' value due to the slightly different AC and IAC, as compared to that used to collect the data in Figures 3 and 4, of the open-coil geometry used for imaging and the differences in heat dissipation incurred by this format.
These observations demonstrate that for printed nanocomposite gels; (i) the initial AC-field heating response (at close-to adiabatic conditions) is largely independent of both GO and FT; (ii) the temperature jumps, T, obtained are similar, in this Fe concentration range, for GO ≥ 0.6 mg mL -1 ; (iii) the FT value largely determines the dissipation and hence the T that can be achieved which sets a lower, Fe-dependent, limit on FT, and most critically; (iv) in our earlier thermography study of BNFs in similar 10 x 10 printed pluronics F127 grids realising a T600 of ~6 °C required an Fe concentration of 140 mg mL -1 [24], compromising applications. Here T600 of ~2.5 °C is observed at only 2.5 mg mL -1 ( Figure   5), at face value this is an efficiency enhancement of 23-fold. PEGylation and nanocompositing with GO contribute to this improvement in the applications relevant format, without compromising printability for pluronics-based hydrogels.

Conclusions
In summary, we investigated the effects GO flakes in the size range of 240 ± 20 nm, on magnetic nanoflower particle suspensions and nanocomposite gels formed with F127. In Strategies of this type will provide improved materials for personalized and autonomous tissue engineering systems, for instance as next-generation spatiotemporally-responsive supports for tissue engineering [40,41].  Magnetic nanoflower (NFs) preparation: Synthesis of iron oxide nanoflowers was based on forced hydrolysis of iron chloride precursors by the polyol route, which was adapted from

Chemicals
Hugounenq et al. [32]. Briefly, in a typical preparation Iron (III) chloride hexahydrate (0.541 g, 2 mmol), Iron (II) chloride tetrahydrate (0.199 g, 1 mmol) were dissolved in a DEG/ NMDEA mixture (37.1 mL, 1:1 v/v) in a 100 mL round bottom flask. A wide range of conditions have been reported [32,44], in selecting this solvent ratio and other conditions we followed the protocol from Hugounenq [32], which provides a detailed analysis of the effect of the key parameters. Separately, sodium hydroxide (0.32 g, 8 mmol) was dissolved in a DEG/NMDEA mixture (19 mL, 1:1 v/v). This solution was then added to the solution of iron chlorides and stirred for 3 hrs. Then, the temperature was increased to 220 °C at 5 °C min -1 by placing the round bottom flask in temperature-controlled heating mantle. The suspension was heated with magnetic stirring for 12 hrs, and then allowed to cool to room temperature.
The black sediments were separated magnetically and washed with an ethanol/ethyl acetate mixture (1:1 v/v) for 3 times to remove impurities. Possible iron hydroxides were removed by treatment with 10% v/v nitric acid (10 mL). Iron (III) nitrate nonahydrate (4.125 g) was dissolved in water (10 mL) and added to the nanoparticles suspension. The resulting mixture was heated to 80 °C for 45 mins to ensure complete oxidation of the nanoparticles, to −Fe2O3. After another treatment with 10% v/v nitric acid, the suspension was washed twice with acetone and diethyl ether and re-dispersed in desired amount of water.

PEG-Gallol synthesis:
This synthesis was adapted from a method by Studart et al. [33]. Poly Scanning electron microscopyelectron dispersive X-ray spectroscopy (SEM-EDX) of gPNF_GO hydrogels: Prior to imaging, the hydrogels were flash-frozen in liquid nitrogen and lyophilized to form freeze-dried solids. These were carefully cut into small porous particles that were coated with a 7.5 nm layer of Iridium using a Quorum Technologies Q150T ES sputter coater. These samples were then imaged using a Hitachi Regulus 8230 scanning electron microscope. EDX was performed using an Oxford Instruments Ultim max 170. Measurements and elemental maps of freeze-dried gPNF and gPNF_GO gels ( Figure   S4) have an associated penetration depth of c.1 μm.
Magnetic hyperthermia: Measurements on bulk homogeneous samples were carried out using a NanoTherics NAN201003 MagneTherm TM AC-field generator (NanoTherics Ltd.; Newcastle-under-Lyme, United Kingdom). The system allows measurement of temperature vs time via a non-metallic OP-Sens optical thermometer to avoid eddy currents. Typically, 1 mL of sample was transferred into a plastic cylindrical shape 2 mL Eppendorf tube with a whole on a cup that fits well the optical thermometer probe. The sample was placed in a thermally insulating polystyrene sample holder to maintain close-to-adiabatic conditions. The deep samples (see supporting information, Table S1). The temperature of the sample was equilibrated in the instrument before the desired field was applied. Unless otherwise noted, measurements were carried out at a frequency of 535 kHz and magnetic field strength of 24 mT. These will be referred to as 'closed system' measurements. For each SAR determination, the error of the initial slope fitting is estimated as standard error calculated using OriginPro, providing average uncertainty of ~4%. We used a fourth order polynomial, increasing that to fifth order had no discernible effect on the linear term. The error in determining the concentration of Fe using AAS is estimated to also be ~2.5%. Combined these give an estimated error in SAR for each measurement of ~5%. This is represented as error bars in Tapping mode) tips. The images were acquired at 512 × 512 pixels resolution at a scan rate between 0.5 and 0.75 Hz and 2 nd order flattened using Nanoscope software ( Figure S2). formulations, all parameter ranges used are summarized in Table 2 below. The gcodes used to print the grids are provided in Table S2. Finally, captured photographs were analysed using ImageJ to establish thickness of lines within printed grids, based on at least 9 measurements across both X and Y grid thicknesses, represented in main text as averages ± standard deviation.

Supporting Information
The Supporting Information is available free of charge at https:// NMR of synthesized PEG-Gallol; Characterization of MNFs; characterization of GO; SEM-EDX characterization of gPNF_GO hydrogels; SAR evaluation of other batches for control GO flakes, sPNF_GO and gPNF_GO; supporting pre-and post-3D printing characterization.

Acknowledgements
We thank Dr. Ian Reid from the Nano Imaging and Material Analysis Centre for supporting SEM imaging.

Funding Sources
The authors acknowledge support from Science Foundation Ireland (

Notes
The authors declare no competing financial interest. All research data supporting this publication are directly available within this publication and associated supporting information.