Dynamic Fractal Ice Nano-Nucleators for Cancer Cryotherapy

Abstract

We present cryo-responsive dynamic fractal ice nano-nucleators (DF-INNs) for cryo-cancer therapy applications. Our development of DF-INNs leverages their structural advantages, which inherently maximizes the number of active sites for heterogeneous ice nucleation. Owing to their radially attached nanocrystalline structure, DF-INNs expose an extensive array of grain boundaries. The rapid precipitation and subsequent radial attachments of nanocrystallites promote the exposure of facets with high Miller indices, intrinsically strained, five-fold twinned nanocrystals, and increasing point defects due to kinetically-limited precipitation. This unique fractal structure culminates in the elevating of freezing temperature compared to Euclidean-shaped ice nucleators. Additionally, the branched fractal structure of DF-INNs facilitates heterogenic ice formation within its nanoconfined region, leading to the production of numerous self-similar small fractal fragments. This fragmentation is primarily driven by nanoconfinement-induced delayed ice nucleation, similar to frost heaving. The shear stress can be easily relieved through grain boundary sliding within the radially stacked DF-INN, making itself prone to cryo-responsive fragmentation. Such dynamic attributes significantly enhance ice nucleating activity, presenting a powerful strategy to increase the efficacy of cryotherapy by enhancing cellular ice formation and in vivo tumor coverage.

Introduction

Cancer cryotherapy, a minimally invasive therapeutic approach that freezes cancer cells by cooling them below nadir temperature (−40°C), has shown promise in treating a wide range of benign and malignant stages of prostate, breast, liver, and lung cancers attributed to unique advantages including minimal invasiveness, short recovery time, the analgesic effect of ice, compatibility with interventional radiology imaging, and preservation of protein’s folded structure, which sets it apart from other heat-based ablation.[1] However, cryotherapy inevitably involves a temperature gradient, leading to the formation of a sublethal temperature zone and supercooled regions in the peripheral area of the treated necrotic cryo-lesion.[2, 3] Cancer cells exposed to these sublethal conditions may not undergo complete necrotic cell death, contributing to a notable recurrence rate.[4, 5]

Various nanoparticles providing heterogeneous nucleation sites offer novel prospects for promoting cellular ice nucleation[6, 7] and addressing the existing challenges in cancer cryotherapy. Their potential role as ice nano-nucleators (INNs) has been exemplified in diverse cryogenic science applications, including cryopreservation, food processing, cloud seeding, and so on.[8–10] The ice nucleation process of INNs primarily entails the phase transformation of supercooled water molecules.[11] Water molecule adsorption, crystal structure matching, reduced interfacial energy, and active site of INNs can enhance the heterogeneous ice nucleation. Various INNs have been studied to promote heterogeneous ice nucleation with their characteristics including excellent lattice matching with ice crystal (AgI), large specific surface area (SiO₂), elevation of thermal conductivity (MgO), and numerous active sites (K-feldspar).[12–14] The recent discovery of ice nucleating active sites such as defects and high energy facets of INNs in supercooled water underscores the importance of INNs nanoscale topography engineering having active sites and maximizing interfacial area.[15, 16] In this context, fractal structures, characterized by self-similar, intricate, and convoluted patterns like blood capillaries, neurons, branched trees, and rivers, have naturally evolved to increase the interfacial area, suggesting branching fractal structures have significant potential for efficient INNs.[17, 18] The maximized surface area of these fractal structured nanoparticles can accommodate promoting water molecules access and reduce critical ice nuclei size, leading to spontaneous ice crystallization.[19] The meticulous development of fractal INNs, customized for specific cryogenic biomedical applications, holds substantial promise for driving advancements in therapeutic outcomes.

Here dynamic fractal ice nano-nucleators (DF-INNs), which can effectively minimize sublethal cryo regions enhance the efficacy of cancer cryotherapy, were synthesized through metal nanocrystal-mediated diffusion-limited aggregation (DLA) process (Scheme 1). Fractal nanoparticles, with their sophisticated, self-similar patterns of metal building blocks, naturally provide a vast specific surface area, and potential active sites for ice nucleation such as numerous, surface defects, high energy facets, and grain boundaries in nanocrystalline structure. DF-INNs demonstrate efficient ice nucleation and cellular ice formation (CIF) during cryo-treatment. In the cryo-thaw cycle of cancer cryotherapy, DF-INNs generate secondary distinct fractal nanoparticles through frost-heaving mediated fragmentation. Subsequently, the cryo-responsive transformation of DF-INNs structure provides an increased number of ice-nucleating active sites, leading to a significant augmentation in heterogeneous CIF. Thus, DF-INNs catalyzed cryo-treatment of cancer cells effectively induces critical damage to cellular membranes. During in vivo validation of the enhanced cancer cell killing of DF-INNs cryotherapy, DF-INNs distribution within the tumor was notably dispersed across the tumor region. As result, DF-INNs cryotherapy effectively eradicated the cancer cells in the ablation zone.

Scheme 1.

Schematic illustration of dynamic fractal ice nano-nucleators (DF-INNs) for enhanced cancer cryotherapy.

Results and Discussion

To enhance ice nucleation in supercooled water during cryotherapy, we synthesized branching DF-INN using the DLA process. This process involves rapid precipitation and aggregation of small metal building block nanocrystals under out-of-equilibrium conditions, resulting in metastable branched fractal nanostructure.[20–23] Unlike polymer-based DLA, where an assembly is more random and less controllable, the attachment of inorganic nanocrystals is governed by the shape and crystallographic orientation of previously attached nanocrystals. This governance allows for more controlled formation of nanocrystal aggregates, thereby enabling precise manipulation of the resulting fractal structure. We demonstrated versatile modulation of branching fractal nanostructures by regulating DLA conditions, including metal cation ratio ([Au³⁺]:[Ag⁺]) and the amount of reducing agent (Fig. S1). Radial attachment of primary building blocks occurs in all synthetic conditions, but the ratio of metal cation affected either two-dimensional ([Au³⁺]:[Ag⁺] = 6.7:1) or three-dimensional fractal structure ([Au³⁺]:[Ag⁺] = 0.9:1). This shift in structural dimensionality could be attributed to a change in the crystal growth mechanism from a layer-by-layer deposition (Frank-van der Merwe growth) to formation of islands (Volmer-Weber growth).[24–26] The amount and concentration of reducing agents affected both interbranch spacing and radial length, determining the dispersion stability of fractal nanostructure in an aqueous solution. Among various morphology of fractal structure, we chose well-confined three-dimensional branching fractal nanoparticles as DF-INNs considering both fractal dimension (D_F), closely associated with the specific surface area for providing active sites for heterogeneous ice nucleation, and interbranch spaces, where water molecules are accommodated. Transmission electron microscopy (TEM) confirmed the branching fractal shape of Au-Ag metallic nanoparticles as outer diameter (257.5 nm), branch thickness (4.4 nm), and interbranch gap (12.7 nm) (Figs. 1b and S2). D_F is a numerical metric that describes random structures in the way these structures fill space, complexity, and its self-similarity, which is different from the description of traditional geometric shapes. Two-dimensional (2D) box-counting method demonstrated the D_F of DF-INN is 1.57 ± 0.03 (Fig. 1c). On the other hand, thermal energy-assisted reduction of gold nanoparticles forms a thermodynamically stable structure with a smooth surface. Thus, we adapted Euclidean-shaped gold nanoparticles (E-GNPs) as a control to investigate the effect of nanoscale topology in ice nucleation (Fig. S3). The specific surface area of synthesized DF-INNs resulted in 35.5 m² g⁻¹ (Fig. 1d), which was 8.2-fold larger than similar sized E-GNPs. High specific surface area of DF-INNs (7.7 ± 0.6°) led to significantly higher hydrophilicity compared to E-GNPs (46.3 ± 11.2°), as observed in the measurement of water droplet contact angle (Fig. 1e), even though both DF-INNs and E-GNPs were functionalized with PEG (3.4 kDa) (Fig. S4). Owing to the hierarchical branching structure compared to the smooth Euclidean shape, DF-INN is advantageous to employ water molecules within interbranch space due to the capillary effect.[27]

Fig. 1.

Synthesis and structural analyses of DF-INNs. a) Schematic diagram of the synthesis of DF-INNs via cholate mediated diffusion-limited aggregation. b) Representative TEM image and the average size of DF-INNs. Data are presented as mean ± s.d. (n = 50). c) Fractal dimension (D_F) of a DF-INN. Scale bars, 100 nm. D_F was measured by box-counting method using Fiji software (n = 10). d) Specific surface area measured by BET using N₂ adsorption-desorption isotherms of the DF-INNs and E-GNPs. e) Contact angle of water droplet on DF-INNs, and E-GNPs-coated glass substrates (n = 5). f) X-ray diffraction (XRD) pattern of DF-INNs. g) HR-TEM image of outermost branch. The image displays the five-fold symmetry of twin boundaries observed at a branching node and a branch tip. Scale bars, 5 nm and 10 nm⁻¹ for real space images and reciprocal space images, respectively. Diffraction peaks and fast Fourier transformation patterns were identified with known references (Au: ICDD card no. 00–004-0784, Ag: ICDD card no. 00–087-0720).

Microstructural analyses revealed that DF-INN is composed of 5.4 nm small Au/Ag crystallites (Fig. 1f). Considering the branch thickness, it indicates small Au/Ag crystallites are one-dimensionally stacked within a single branch rather than lateral attachment. Also, fast Fourier transformation patterns from neighboring small crystallites confirmed that there is no specific crystallographic order within stacking branches, as a characteristic feature of diffusion-limited aggregates (Fig. S5).[28] Inductively-coupled plasma mass spectroscopy (ICP-MS) revealed that the atomic ratio is [Au]:[Ag] = 26.9:1. Considering the initial injected precursor ratio ([Au⁺]:[Ag⁺] = 0.9:1), it is evident that the final DF-INN is predominantly composed of gold. Interestingly, five-fold twinned, decahedral building blocks were observed at the branching nodes and branching tips (Fig. 1g). Face-centered cubic noble metals like Au and Ag have various polyhedral Wulff shapes including five-fold twinned decahedron determined by the surface energy ratio of exposed facets.[29] The DLA of inorganic nanocrystals differs markedly from that of soft materials, primarily due to the influence of the crystalline substrate and facet orientation of building block nanocrystals, which directs the orientation and shape of the resulting fractal structure. This influence ensures that the initial pool of building block nanocrystals significantly determines the emergent the self-similar pattern. Probabilistic nucleation and stochastic attachment on five-fold twinned building blocks played a crucial role in guiding the overall fractal pattern. Five-fold twinned nanocrystals not only have relatively high energy due to intrinsic strain to maintain the symmetry but also contribute to assigning a number of grain boundaries along branches.[30] When the surface of metal is sufficiently capped with surfactants or other adsorbates, multifaceted Au nanoparticles with high-index and low-index facets were experimentally observed.[31, 32] In case of DF-INNs, cholate acted as a transient passivating agent during the nucleation and attachment of primary crystallites. Altogether, DF-INN, characterized by their high specific surface area in fractal pattern, nanocrystalline microstructure, and multifaceted building blocks, holds a great potential for catalyzing the nucleation of ice in supercooled solution by locally concentrating H₂O molecules on its surface.

Cancer cryotherapy uses extremely cold cryogen (liquid N₂ or Ar gas), thus the criteria for INNs shift towards the ability to reduce the degree of supercooling under rapid cooling. The enhanced interfacial interaction between DF-INNs and water molecules directly contributed to the increase in freezing temperature (T_F). In differential scanning calorimetry (DSC) measurement, cooling curves of DF-INNs aqueous solution demonstrated a significantly elevated T_F (−17.3°C) by about 9.1°C and 4.2°C compared to water (−25.9°C) and E-GNP aqueous solution (−21.5°C), respectively (Fig. 2a). DF-INNs providing heterogeneous ice nucleation sites successfully mitigated the degree of supercooling in the solution, resulting in increasing the ice nucleating temperature.[33, 34] In addition to the primary exothermic T_F peak, a delayed freezing peak was observed at −31.5°C. It indicates temporally separated two different ice nucleation by primary active sites near branching nodes and the nano-confined region in which T_F is lowered by Gibbs-Thomson effect.[35] The Gibbs-Thomson effect explains that the degree of freezing point depression is inversely proportional to critical radius (interbranch gap). Therefore, within the nano-confined spaces between the internal branches of DF-INNs, water pockets can remain unfrozen until lower temperature, resulting in delayed ice crystallization. Although there was some variation in the degree of delayed freezing peak position in DSC originating from structural deviation in DLA aggregates, the overall enthalpy change remained relatively consistent as approximately 0.67% of the enthalpy change in primary freezing. DF-INNs treated with an additional cryo-cycle showed increased T_F (from −17.3°C to −16.2°C and −14.6°C after one and two freeze-thaw cycles, respectively) compared to E-GNPs (non-significant difference, from −21.5°C to −21.1°C and −21.1°C after one and two freeze-thaw cycles, respectively) or pure water (−25.9°C) in DSC measurement (Fig. 2b). DF-INNs exhibited superior T_F elevation in rapid cooling conditions, surpassing the performance of conventional INNs such as MgO and AgI (Table S1). Cryo-responsive DF-INNs stand out by demonstrating a remarkable and additional T_F elevation upon freeze-thaw cycles, thereby substantially decreasing the extent of supercooling efficiently. This attribute makes DF-INNs particularly suited for cryo-sensitizer. The significant ice nucleating activity of DF-INNs can be attributed to their unique surface structure. Unlike conventional INNs with smooth surface, the irregular surface topology of DF-INNs, including high index facets, surface defects, and grain boundaries increases the number of potential nucleation sites. This enhanced nucleation capability allows DF-INNs to elevate T_F even at much lower concentration of INNs. TEM images clearly demonstrated DF-INNs exhibited significant structural fragmentation after freeze-thaw cycles (Fig. 2c). Generated small fragments of DF-INNs had a similar periodic branching fractal nanostructure (D_F: 1.84) with original DF-INNs (D_F: 1.57) (Fig. 2d). An increase of D_F suggested the morphologies of cryo-treated DF-INNs became more compact and efficient in filling space, making them enlarged exposed surface area from original DF-INNs (Fig. S6).[36] It also contributed to the T_F elevation by increasing the number of active sites for heterogeneous ice crystallization as structural fragmentation of DF-INNs. Although the self-similarity of objects derived from their 2D projections has some discrepancies compared to the actual 3D objects, there is a significant correlation. This correlation provides a feasible method to understand and track changes in the object, even when it is technically hard to utilize accurate 3D TEM images.[37] There are Hydrodynamic size changes (from 215.1 to 5.9 nm after Cryo2) further confirmed cryo-responsive dynamic structural change of DF-INNs (Fig. 2e). This result indicated the number of small fragments increased due to the cryo-responsive transformation, and relatively large broken pieces neglected considering the DLS intensity proportional to the 6^th power of radius of nanoparticle. The emergence of an extinction peak around 440 nm of DF-INNs after cryotreatment was also observed. We demonstrated that the release of small fragments by the cryo-responsive fracture led to an increased in exposed sections and a decrease in light scattering from the dense branched structure. Then, in situ mechanical stress loading to a DF-INN using a nano-indenter depicted the possible fracture of branches in DF-INNs by the shear force of ice nucleation process (Fig. 2g). Stress-strain curve showed a tumbled region, implying nucleation and propagation of dislocations to relieve mechanical loading, consequently resulting in partial failure of branches in a range of 100 to 500 MPa mechanical stress (Fig. 2h). The observed structural deformation of DF-INNs entails internal ice nucleation within the DF-INNs, followed by sequential volumetric expansion exerting cumulative ice crystallization pressure on the surrounding branches (Fig. 2i). The increased curvature and interfacial energy led to the delayed ice nucleation, separated from the primary freezing events at active sites. Then, the secondary ice formation including approximately 9% volumetric expansion, with the exterior of DF-INNs frozen, applied cumulative crystallization stress to surrounding branches of DF-INN. Therefore, the shear stress could be accumulated at the grain boundaries of thin branches, causing the sliding of stacked building blocks and subsequent fragmentation.[38, 39] Termed as the “nano-frost heaving” effect, this phenomenon underscores the dynamic interplay between ice formation and mechanical stress within the nanostructured framework.

Fig. 2.

Freezing temperature elevation and cryo-responsive transformation of DF-INNs. a) Cooling DSC curves of pure H₂O, E-GNP, and DF-INN solution. The concentration of each nanoparticle is 0.2 mg mL⁻¹ in H₂O. Magnified graph on the right presents the secondary freezing peak in DF-INN solution around −32°C (blue-dotted circle in the left graph). Table summarize the freezing temperature (T_F) and the enthalpy change of each sample. b) T_F of H₂O, E-GNP, and DF-INN solution (0.2 mg mL⁻¹) upon cryotreatment (n = 3). c) Representative TEM images of DF-INN fragments after two consecutive freeze-thaw cycles (Cryo2; one cryo cycle comprised of freezing (−80°C for 3 min) and thawing (37°C for 5 min)). d) D_F of pristine DF-INNs, broken DF-INNs, and small fragments. Scale bars, 100 nm. Fractal dimension was calculated from TEM images using a box-counting method (n = 10). e) Hydrodynamic size distribution and f) UV-visible light extinction spectra of DF-INN upon cryotreatment. g) Representative snapshot SEM images during the in situ mechanical testing of a DF-INN in SEM-FIB manipulating system; i. before deformation, ii. during deformation, and iii. complete fracture. Scale bars, 500 nm. h) Stress-strain curve of DF-INN under mechanical load. As a DF-INN was subjected to compressive stress, more branches supported the structure, resulting in a tumbled stress-strain curve. i) Schematic diagram of nano-frost heaving mediated structural deformation of DF-INN. All data are presented as the mean ± s.d. P-values: *p<0.05, and **p<0.01. Statistical significance was evaluated by an unpaired two-tailed t-test in b).

The capability of DF-INNs to enhance ice nucleation and increase T_F was further demonstrated by the time-dependent probability of CIF in a cryo-chamber (Fig. 3a). As cell-seeded glasses were washed before exposure to cooling down, there was no significant difference in the freezing of extracellular solution across all samples, indicating that the cryo-stage controlled the cooling rate consistently. However, DF-INNs treatment accelerated the onset of CIF by up to 60 s. The CIF ramping rate was also enhanced by 1.5 and 1.7 times compared to E-GNPs treated cells or non-treated cells, respectively. Effective DF-INNs mediated CIF injures the integrity of cellular organelle and membranes, directly leading to cell death during cryo-treatment. DF-INNs treated cells in both single and double freeze-thaw cycles exhibited enhanced cancer cell killing effects, according to DF-INNs concentration (Fig. 3b). At 0.5 mg mL⁻¹, comparative cancer cell killing results showed 22.5% vs. 32.3 vs. 59.3% (Cryo only vs. E-GNPs+Cryo vs. DF-INNs+Cryo) (1^st cycle) and 38.3% vs. 51.6% vs. 82.0% (2^nd cycle). DF-INNs and their dynamic fragments generation during two cycles of cryotreatment appear to contribute to significantly improved cancer cell necrosis over a single cryo-cycle. Based on these results, two cryo-cycles were used with DF-INNs for subsequent investigation. Representative confocal microscope images of DF-INNs-cryo-treated cells demonstrated severe damages to cellular membrane and lysosomal structures, while shapes of E-GNPs-cryo-treated cells were maintained with consistent cellular phenotype even after cryotreatment (Fig. 3c). Interestingly, colocalization between lysosome and DF-INN significantly decreased after cryotreatment, indicating DF-INN escaped from the lysosome. It suggested that DF-INN have a potential to maximize intracellular damage than E-GNPs. Bio-TEM measurement further visualized DF-INNs cryotreatment mediated cell death and DF-INN fragments. As demonstrated in confocal images, DF-INNs cryotreatment caused substantial cell membrane damage, whereas DF-INNs treated cells without cryotreatment showed no apparent membrane damage (Fig. 3d). Additionally, a large number of small fragments of DF-INNs were conspicuously found near the cell membrane debris. However, only DF-INNs treated cells without cryotreatment showed a preserved original shape of DF-INNs with minor dissociated DF-INNs in endosomes. Cellular DF-INNs might undergo cryo-responsive nano-frost heaving fragmentation and diffuse out from the dead cells. Observed significant damages of cellular membrane and organelles are likely contributing to various cell death mechanism. Thus, we investigated the dominant cell death mechanism after cryotreatment using Annexin V and propidium iodide (PI) assay. As qualitatively confirmed above, DF-INNs-cryo-treated cells exhibited significantly increased necrosis after two consecutive freeze-thaw cycles, with 85.7% necrotic cells compared to other groups (CTRL; 29.5%, E-GNP; 45.3%) (Fig. 3e). Notably, after the second freeze-thaw cycle, the proportion of necrotic cells (AnnexinV⁺PI⁺) in DF-INNs-cryo-treated group surged by 2.84 times compared to their levels after a single freeze-thaw cycle, whereas the E-GNPs-cryo-treated group showed a 1.88-fold increase from the single cycle. This marked difference is attributed to the release of cryo-responsive fragments during the thawing process in the initial freeze-thaw cycle and the subsequent increase in active sites during the second cooling phase. It highlights that the effectiveness of DF-INNs enhances critical cellular damage during the cryotreatment, as DF-INNs prolonged the exposure time to nadir temperature by elevating T_F. This cryo-responsiveness of DF-INNs underscores their potential as cryo-sensitizer for cancer therapy to maximize therapeutic outcomes by manipulating the thermal dynamics during cryotherapy.

Fig. 3.

Cellular ice formation and enhanced cancer cell death of DF-INNs. a) Representative optical microscopy images of TrampC1 cells under cryo-treatment after incubation with control (CTRL, without nanoparticle treatment), E-GNP, or DF-INN for 4 h. Cells lost light transmission was identified as frozen cells (blue arrows), compared to viable cells (yellow arrows). Scale bars, 2.5 μm. Time-dependent probability of cellular ice formation (CIF) upon cryotreatment. In images with more than 150 viable cells, the fraction of frozen cells was counted at 30 s intervals while cooling (n = 5). b) Cytotoxicity of E-GNP and DF-INN against TrampC1 cells using CCK-8 agents with cryotreatment. Each cryotreatment comprises freezing (−80°C for 3 min) and thawing (37°C for 5 min), which indicates a “Cryo”. Control (CTRL) means without nanoparticle treatment and non-treatment (N.T.) means incubation at 37°C (n = 6). c) Confocal microscopy images of E-GNP, or DF-INN-treated TrampC1 cells before and after cryotreatment. Each color indicated nucleus (blue), membrane (red), lysosome (green), and E-GNPs or DF-INNs (magenta), respectively. Scale bars, 10 μm. Pearson’s correlation coefficient between two fluorescent sources (lysosome and DF-INNs) upon cryotreatment (n = 3). d) Representative Bio-TEM images of TrampC1 cells with DF-INNs before and after cryotreatment. Images in the lower row are enlarged images of the boxed part of images in the upper row. Arrows indicate broken DF-INNs. Those images indicate the potential of DF-INNs to damage intracellular organelles by enhancing CIF. e) Necrotic cell death (Annexin-V⁺PI⁺) detected by Annexin-V/PI assay (n = 3). All data are presented as the mean ± s.d. P-values: ns, not significant, *p<0.05, **p<0.01, and ***p<0.001. Statistical significance was evaluated by an unpaired two-tailed t-test in a, b, c, and e.

DF-INNs, incorporating heavy atoms (Au and Ag), exhibit radiopacity on computed tomography (CT), facilitating their use in image-guided delivery and the monitoring of cryo-responsive distribution within tumors (Fig. S7). Initial CT scans revealed i.t. injected cryo-sensitizers (DF-INNs or E-GNPs) were successfully localized at the center of tumor prior to cryotherapy, with average Hounsfield Unit (HU) around 1743, clearly distinguishable from neighboring tissue (Fig. 4a, and 4b). Volumetric CT contrast of cryo-sensitizers had been monitored for 14 days post-cryotherapy to evaluate cryo-responsive intratumoral distribution changes of DF-INNs and E-GNPs. Volumetric CT contrasts from DF-INNs or E-GNPs were monitored by filtering out the CT contrast below 500 HU to exclude noise signals from neighboring tissue (Figs. 4c, 4d, and S8). The reconstructed CT volume was 3.3-fold expanded after 3 days of cryotherapy in DF-INNs group, while DF-INNs in non-cryo-treated group kept a monotonous aggregate at the initial injection site (1.5-fold) (Fig. 4e).[23, 40] However, E-GNPs treated group didn’t have significant difference until 3 days after the cryotherapy, 1.3-fold vs. 1.4-fold (E-GNPs vs. E-GNPs+Cryo). This exceptional expansion of DF-INN occupied volume in tumor after cryotherapy could be attributed to cryo-induced cellular destruction, cryo-responsive fracture of DF-INNs and fluid dynamics within the tumor tissue. Furthermore, hematoxylin-eosin (H&E) staining of tumor tissues demonstrated diffused DF-INNs fragments within the tumor with their cryo-responsive fragmentation (Fig. 4f). As the size of cryo-sensitizers decreases upon cryo-responsive fragmentation, small fragments had been diffused out toward the peripheral tumor region. Severe cryo-mediated cellular injuries involved necrotic cell death leading to a discharge of intracellular components, resulting in loosened tissue density, which might disperse small DF-INNs away from the center. Consequently damaged tissue underwent transition into an edematous tissue, and serous drainage or blood serum in the cryo-damaged tumor tissues undergoes active in-out flow.[41–43] We observed a drastic accumulation of fluid surrounding the tumor and adjacent subcutaneous space in T₂-weighted magnetic resonance imaging (MRI) following cryotherapy (Fig. S9). This finding indicates cryotherapy induces considerable cell death, leading to tissue edema and active fluid exchange within the cryo-damaged tissues. We postulate that the dispersion of DF-INNs is facilitated by the changes in interstitial fluid dynamics and a decrease in tumor density. After cryotherapy, DF-INNs’ fragments reached the peripheral tumor regions at post-cryotherapy. Consequently, the enhanced distribution of DF-INNs promotes broader tumor coverage, allowing a more extensive area of the tumor to be effectively treated by cryotherapy.

Fig. 4.

In vivo cryo-responsive DF-INNs distribution toward peripheral region upon cryotherapy. a, b) In vivo monitoring of intratumoral DF-INNs (a) or E-GNPs (b) distribution via coronal view of CT scan for 7 days after cryotherapy. Scale bars, 5 mm. c, d) Representative three-dimensional (3D) reconstructed DF-INNs (c) or E-GNPs (d) volume for 7 days after cryotherapy. Each 3D model represented the DF-INNs or E-GNPs occupied volume having a Hounsfield unit above 500 through the 3D slicer software. The black dotted lines indicate the virtual border of the tumor tissue and were estimated by measuring the tumor width and length with a caliper at each imaging time point. Relative nanoparticle (either DF-INN or E-GNP) occupied volume change upon cryotreatment for 7 days (n = 4). e) H&E-stained sections of mouse tumor tissues. White dotted lines indicate the tumor boundary between the tumor and normal tissues. Scale bars, 50 μm. All data are presented as the mean ± s.d. P-values: ns, not significant, *p<0.05, **p<0.01, and ***p<0.001. Statistical significance was evaluated by an unpaired two-tailed t-test in e.

Like other heat-based ablation, cryotherapy fundamentally relies on heat transfer, a process significantly affected by the tumor microenvironment’s inherent heterogeneity, such as size, location, and adjacent vascular structure.[44] As a result, some cancer cells may receive insufficient cold stress, enabling them to withstand the cryotreatment and ultimately leading to recurrence. To determine the impact of enhanced CIF and the distribution of intratumoral DF-INN on therapeutic efficacy, particularly in minimizing tumor cells suffering non-critical cold stress, we introduced TrampC1 tumor-bearing mice model for the cancer cryotherapy (Fig. 5a). Following the clinical regimen, we designed a cryotherapy procedure of two consecutive freeze-thaw cycles, where the tumor cooled below nadir temperature (−30°C) for 1 min and passively thawed (Fig. S10). Tumor volume had been monitored for 2 weeks after cryotherapy (Fig. 5b). Approximately one week after the cryotherapy, we observed a slight increase in the tumor volume in both E-GNPs or DF-INNs treated groups, which is attributed to the formation of edema. After edema formation, the healing process, which involves removing necrotic cells, leads to decrease in the inflated volume. In vivo DF-INNs-assisted cryotherapy of the TrampC1 tumor effectively regressed 69.2% of the initial tumor. In a comparison with non-fractal E-GNPs+Cryo, DF-INNs’ catalytic effect showed a 1.6-fold increased tumor regression in 2 weeks. 14 days after cryotherapy, we acquired tumor tissue and stained with H&E, Ki-67 (cancer cell proliferation marker), and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL, apoptotic or necrotic cell death marker) to confirm the enhanced cryotherapy margin using cryo-responsive DF-INNs (Fig. 5c). Both E-GNPs and DF-INNs cryotherapy groups displayed cryotherapy-mediated cell death in H&E and TUNEL-stained tissue at the center of the tumor (Fig. S11). There were no TUNEL stained cells in the DF-INNs treated group, indicating DF-INN mediated cryotherapy significantly induced extensive necrosis and inhibition of tumor growth. However, E-GNPs and DF-INNs mediated cryotherapy showed significant effects on the differences in the tumor’s peripheral region, as seen in the magnified images (Fig. 5c). Those magnified regions in the second column appear to act as bumps to block the propagation of critical cold stress, as can be observed in the whole tumor images in the first column. Cancer cells in these obstructed areas likely receive insufficient cold stress due to their lower thermal conductivity relative to surrounding areas. Consequently, the areas treated with E-GNPs cryotherapy displayed a significantly higher proportion of Ki-67 positive cells (16.8 ± 2.5%), indicating an active state of proliferation status even after cryotherapy (Figs. 5c and 5d). In contrast, the DF-INNs cryotherapy groups showed a substantially lower proportion of Ki-67 positive cells (2.3 ± 1.9%). These results highlighted that enhanced CIF and expanded tumor coverage achieved with DF-INNs effectively ablated tumor cells and inhibited further tumor proliferation. The enhanced tumor regression might be attributed to combined DF-INNs mediated cryo-cell injury and sequential CD8⁺ T cell activation. Tumor infiltrated CD8⁺ T cells were 1.6-fold and 1.4-fold increased than control and E-GNPs cryotherapy, respectively (Fig. 5e). The activation of the adaptive immune system, evidenced by increased tumor infiltration of CD8⁺ T cells following cryotherapy, could be sensitized by the release of tumor-associated antigens and, immunestimulatory molecules such as cytokines, and damage-associated molecular patterns (DAMPs). At the same time, DF-INNs cryotherapy exhibited a 1.5-fold higher PD-L1 expression level than cryotherapy only (Fig. 5f). PD-L1 expression might be upregulated by tumor cells exposed to the antitumor immune system undergoing DNA damage and activating type Ⅰ interferon (IFN) signaling pathway.[45, 46] Observed results proved that DF-INNs’ cryo-responsive fragmentation enhanced irreversible cancer cell death efficiently. Further, the immune activation and PD-L1 overexpression by DF-INNs+Cryo provided a clue for an effective and combinational cryo-immunotherapy using PD-L1 blockade.

Fig. 5.

In vivo cryo-responsive DF-INNs distribution and tumor responses. a) Schematic illustration of enhanced cryotherapy mediated by cryo-responsive DF-INNs. The actual temperature in the cryotherapy zone might be lower than the temperature displayed by the infrared camera, which has a lower limit of detection is −10°C. The IR camera having a temperature detection limit was used for visualization of relative temperature distribution of the therapeutic area. Please refer to Fig. S10 for the actual temperature during the cryotherapy. b) Individual tumor growth trend was monitored for 14 days. Final tumor volume after cryotherapy on day 14 (n = 4). c) H&E and Ki-67-stained tumor images display enlarged cryotherapy-coverage area in the CTRL, E-GNPs, or DF-INNs cryotherapy group. Scale bars, 1 mm (whole tumor) and 100 μm (magnified images). d) Proportion of Ki-67 positive cells in the region exposed to insufficient cold stress (magnified part in c is one example from each sample, n = 5). e) Infiltrated cytotoxic T cells (CD3⁺CD8⁺) in tumor on day 14 (n = 6). f) Increased PD-L1 expression in tumor cells after 14 days post-cryotherapy (n = 6). All data are presented as the mean ± s.d. P-values: ns, not significant, **p<0.01, and ***p<0.001. Statistical significance was evaluated by an unpaired two-tailed t-test in b, d, e, and f.

Conclusions

Cancer cryotherapy stands as a highly effective approach for precision image-guided therapy, demonstrating notable reductions in complications, pain, costs, and recovery time. Ice nucleating nanomaterials have emerged as pivotal contributors to the expansion of cancer cryotherapy. This study introduces engineered DF-INNs that enhance cryotherapy by tailoring the branching fractal nanostructure via DLA. DF-INNs significantly elevate T_F originated from their large number of heterogeneous ice nucleating active sites, which include high energy surfaces, grain boundaries, surface defects, and a substantial specific surface area. Additionally, the “nano-frost heaving” effect generates self-similar ice nucleating fragments. Integrating DF-INNs with cryotreatment significantly increased necrotic cell death through DF-INNs mediated CIF. Dynamic fluid exchange after cryotherapy dispersed small fragments throughout the entire tumor region. DF-INNs induce elevated levels of tumor-associated antigens, cytokines, chemokines, and DAMPs, marking a notable advancement in circumventing the constraints associated with conventional cancer cryotherapy. This offers promising translational prospects for multi-functional ice nano-nucleators in the context of image-guided cancer cryotherapy.

Experimental section

Synthesis of dynamic fractal Au nanoparticles

We prepared 100 mL of 2 mM sodium cholate hydrate (C₂₄H₃₉NaO₅∙xH₂O, Sigma-Aldrich, MO, USA) solution dissolved in deionized water (H₂O, Millipore Direct-Q UV 3). 10 mL of 1.35 mM gold (III) chloride trihydrate (HAuCl₄∙3H₂O, Sigma-Aldrich, MO, USA), 1.5 mL of 10 mM silver nitrate (AgNO₃, Thermofisher, MA, USA), and 1.5 mL of 0.16 M L-ascorbic acid (C₆H₈O₆, Sigma-Aldrich, MO, USA) were injected to sodium cholate solution with 1 min of interval under magnetic stirring. The turbid solution turned into dark grey color right after adding the L-ascorbic acid to the solution and 20 s after the injection, the magnetic stirring was stopped, and the solution was kept stable for 1 h. Then, 50 mg of HS-PEG-COOH (HE003017–3.4K, Biopharma PEG, MA, USA) was added to the solution and magnetically stirred for 24 h. The final PEGylated DF-INNs were purified by centrifugation with H₂O 5 times to remove excessive cholate and PEG. DF-INNs were re-dispersed by pipetting without sonication after each cycle of centrifugation.

Cryo-responsive fracture of DF-INNs

To trigger cryo-responsive fracture, 1 mL of DF-INN solution (0.2 mg mL⁻¹) in a microtube was cooled with a rapid cooling rate (20°C min⁻¹) for 3 min. Then, the microtube was embedded in a bath filled with heated metal beads which were set to 37°C for 5 min. To avoid any confusions in applying cold stress toward NPs, cells or animals, we defined one freeze-thaw cycle, and two consecutive freeze-thaw cycles as Cryo1 and Cryo2, respectively. Samples treated under −80°C for 3 min during the freezing, and passively thawed under 37°C for 3 min.

Morphological-, structural-, surface characterization of DF-INNs

The morphology of DF-INNs was observed using TEM (FEI Tecnai Spirit G2, FEI Company, OR, USA) at 80 kV of accelerating voltage. The diluted NP solution (below 0.05 mg mL⁻¹) was dried on a carbon-coated Cu grid. We measured the outer diameter of DF-INNs (n > 50), branch thickness (n > 100, 10 data points were acquired from a single NP), and interbranch gap (n > 100, 10 data points were acquired from a single NP) using Fiji (NIH, USA). The hierarchical fractal structure was further investigated by double C_s and monochromated TEM (TITAN, FEI Company, OR, USA). For the high-resolution TEM (HR-TEM) analysis, NP solution was dried on either a lacey carbon grid or an ultrathin carbon grid after removing formvar using chloroform. Nitrogen adsorption isotherms of DF-INNs and E-GNPs were obtained under the pressure range of P P₀⁻¹ from 0.05 to 0.2. using a NOVA Surface Area Analyzer (Nova 2200e, Quantachrome instrument). Before measurements, samples were degassed for 12 h at 300°C. The pore size, surface area, and pore volume were analyzed through Brunauer–Emmett–Teller (BET) and Barrett-Joyner-Halenda (BJH) algorithm. The hydrophilicity of each NP was measured with a contact angle measurement system equipped with a digital camera (OCA200, Materials Characterization Instruments Korea, Korea). A concentration of 5 mg mL⁻¹ of DF-INN-/ E-GNP- was densely coated on the hydrophobic glass substrate to monitor alterations in contact behavior. Subsequently, A 1 μL droplet of deionized water was meticulously dispensed onto the NP-coated surface using a microsyringe, and the static contact angle was precisely evaluated with software FIJI (NIH, USA). Fourier transform infrared (FT-IR) spectroscopy of DF-INNs and E-GNPs before and after PEGylation was measured by mounting 1 mg dry powder of each sample on an attenuated total reflectance (ATR) stage equipped with diamond crystal window. To characterize crystalline phase of NPs, XRD analysis (Malvern PANalytical, Malvern, United Kingdom) was conducted with a Cu Kα radiation (λ=1.5406 Å). We attained XRD patterns from 20° to 80° in a θ−2θ geometry. Based on the international center for diffraction data (ICDD) of Au (00–004-0784) and Ag (00–087-0720), diffraction peaks were assigned with crystalline index of Au and Ag. Atomic quantification of Au and Ag element in DF-INN was measured using ICP-MS (Thermo iCAP Q). Each sample was measured after digestion using nitric acid, hydrogen peroxide and hydrochloric acid at 70°C for 6 hours. DSC 250 (TA instrument, DE, USA) was used to measure the effect of NPs on the T_F of aqueous solution. The system was turned on more than 30 min before the experiment. The protective N₂ gas was flowed with a rate of 50 mL min⁻¹. The samples of pure H₂O, DF-INN (0.2 mg mL⁻¹), and E-GNP (0.2 mg mL⁻¹) were embedded in the crucible (Tzero Aluminium, TA instrument, DE, USA). The scanning sequence was set as follows: i) equilibrium to 10°C, ii) Cooling from 10°C to −50°C at a constant cooling rate of −20°C min⁻¹. The freezing point, and the change of enthalpy were calculated using the TRIOS software (TA instrument, DE, USA). DSC measurement was repeated at least 3 times with different batches of DF-INN to validate the reliability and stability of the cooling curve of DF-INNs, with particular emphasis on the secondary freezing phenomenon. The hydrodynamic size of samples was measured by Zetasizer Nano ZS90 (Malvern PANalytical, Malvern, UK). Briefly, 1 mL of diluted sample (0.01 mg mL⁻¹) in a transparent disposable cuvette was irradiated 633 nm He-Ne laser. UV-vis extinction spectra of DF-INN solution with non-treated, Cryo1, and Cryo2 was measured with spectrophotometer (Synergy HT, BioTek, VT, USA). DF-INNs were spread on the SiO₂ substrate and dried at vacuum for 48 h. Then, the glass substrate was inserted into the dual scanning electron microscopy-focused ion beam system (SEM-FIB, Quanta 3D FEG, FEI Company, OR, USA). The stress-strain curve was calculated from the recorded load-displacement data.

Synthesis of Euclidean shape Au NPs

We synthesized E-GNPs (around 200 nm in diameter) using seed-mediated growth.[47] Briefly, 10 nm NPs were synthesized by heating up the solution composed of 20 mM HAuCl₄∙3H₂O 10 mL and 90 mL of H₂O to 100°C for 30 min under vigorous magnetic stirring. Then, 10 mL of 40 mM of trisodium citrate solution (Na₃-cit, HOC(COONa)(CH₂COONa)₂·2H₂O, Sigma-Aldrich, MO, USA) was added to the solution and mixed till the mixture solution turn to dark red color. Then, 5 mL of NP solution were dispersed in 65 mL of H₂O. Then, the solution was heated up to 90°C under magnetic stirring. 0.6 mL of 60 mM Na₃-cit solution and 0.3 mL of 20 mM HAuCl₄ were periodically added to the solution with 30 min interval. After solution turn bright pink color, the reaction was stopped, and NPs were purified with H₂O. Finally, 20 mg of E-GNPs were dispersed in 10 mL H₂O and functionalized with 10 mg of HS-PEG-COOH for 24 h and washed by H₂O.

Calculation of fractal dimension of DF-INNs

The fractal dimension D_F of NP was calculated via a classic 2D box-counting method using the Fraclac plugin in Fiji (NIH, USA).[48] D_F was calculated by the following equation; ${\mathrm{D}}_{\mathrm{F}}={\mathrm{l}\mathrm{i}\mathrm{m}}_{\mathrm{\epsilon }\to 0}\frac{\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{N}\left(\mathrm{\epsilon }\right)}{\mathrm{l}\mathrm{o}\mathrm{g}(\frac{1}{\mathrm{\epsilon }})}$, where $\mathrm{N}\left(\mathrm{\epsilon }\right)$ is the number of boxes of side length $\mathrm{\epsilon }$ required to cover the DF-INN silhouette. In our system, $\mathrm{\epsilon }$ was in a range from 293.50 nm to 0.16 nm. ${\mathrm{D}}_{\mathrm{F}}$ calculation was repeated for 12 times in different orientations in the same image. Then, the average ${\mathrm{D}}_{\mathrm{F}}$ of pristine DF-INNs, broken DF-INNs, and small fragments were derived from 10 different images.

Cell culture

TrampC1 (mouse prostate cancer cell line, CRL-2730) was purchased from American Type Culture Collection (ATCC). Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) high glucose supplemented with 5% heat inactivated fetal bovine serum (FBS), 5% Nu-serum IV, 0.005 mg mL⁻¹ bovine insulin, 10 nM dehydroisoandrosterone, and 1% penicillin/streptomycin (Gibco, Thermofisher, MA, USA). Cells were cultured at 37°C with 100% humidity and 5% CO₂ and subcultured in new media every 1–2 days.

Cytotoxicity assay

TrampC1 cells (1 × 10⁴ cells well⁻¹) were seeded in 96-well culture plates. After 24 h incubation, the cells were treated with E-GNPs or DF-INNs (0.05, 0.1, 0.2, and 0.5 mg mL⁻¹) for 4 h, and cryotreated the indicated number of times. After 24 h incubation, cell viability was quantified by the colorimetric assay using CCK-8 reagents (Dojindo Inc., MD, USA) at 460 nm by spectrophotometer (Synergy HT, BioTek, VT, USA). For the AnnexinV-PI assay, TrampC1 cells (3 × 10⁵ cells well⁻¹) were seeded in 12-well culture plates for 24 h, incubated with E-GNPs or DF-INNs (0.2 mg mL⁻¹) for 4 h, and cryotreated the indicated number of cycles. Then, cells were stained with FITC-Annexin-V and propidium iodide (PI) according to the manufacturer’s instructions (Abcam, Cambridge, UK). The results were examined by flow cytometry and quantified using FlowJo software (TreeStar) (BD Bioscience, NJ, USA).

Confocal laser scanning microscopy

TrampC1 cells (1 × 10⁵ cells well⁻¹) were seeded on a glass bottom dish (Nunc^™, Thermo Scientific, MA, USA) at least 16 h before use (MatTek, MA, USA). 10 mg of E-GNPs and DF-INNs in 10 mL of H₂O (1 mg mL⁻¹) were tagged with amine-terminated Cy5.5 (Lumiprobe, MD, USA) via N-ethyl-N′-(3-(dimethylaminopropyl)carbodiimide) (EDC, 10 mM, 1 mL)-N-hydroxysuccinimide (NHS, 10 mM, 1 mL) reaction for 3 h, and subsequently washed with fresh deionized water. Then, Cy5.5-amine dye solution (0.2 mg mL⁻¹) was added to each glass vial for 24 h under gentle shaking. After fluorescent dyes conjugated to the surface of NPs, excessive dyes were washed out with deionized water. Attached cells were co-incubated with Cy5.5-tagged E-GNPs and Cy5.5-tagged DF-INNs (0.2 mg mL⁻¹) for 4h, performed twice cryo-treatment in −80°C for 3 min each cycle, stained with lysosome (ab445469, Abcam, Cambridge, UK), Tetramethyl indocarbocyanine Perchlorate (DiIC 18, D282, Invitrogen, Thermofisher, MA, USA) for cell membrane, and DAPI (Invitrogen, Thermofisher, MA, USA) for visualize nucleus, and fixed with 4% paraformaldehyde. Images were taken with a Nikon A1R spectral confocal laser scanning system on an inverted microscope (Nikon, Tokyo, Japan) with an argon-krypton laser as excitation source. Image analyses were quantified with Fiji (NIH, USA). To investigate the colocalization of each cellular component, we calculated Pearson correlation coefficient ($r$) between two fluorescent dyes; lysotracker-Cy5.5 in Fiji from the monochromatically split images using the following equation; $r=\frac{{\sum }_i\left(A_i-\bar{A}\right)\left(B_i-\bar{B}\right)}{\sqrt{{\sum }_i{\left(A_i-\bar{A}\right)}^2{\left(B_i-\bar{B}\right)}^2}}$, where ${\mathrm{A}}_i$, ${\mathrm{B}}_i$, $\bar{A}$, and $\bar{B}$ refers to the intensity of fluorophore A and B channels of pixel $i$, an average fluorescent intensity of channel A and B, respectively.[49] The colocalization coefficient was averaged from three different images.

Cell pellet preparation for TEM observation

TrampC1 cells (3×10⁵ cells well⁻¹) were seeded in a 35 mm diameter culture dish for overnight. When they fully adhered to the substrate, we replaced the growth media with DF-INNs-containing growth media (0.2 mg mL⁻¹). After 4 h, we incubated a culture dish under −80°C for 3 min to induce cryo-responsive injury. Cell culture dish passively thawed under room temperature. The remaining TrampC1 cells on a glass substrate were detached by adding trypsin-EDTA and collected as a cell pellet by centrifugation. The cell pellet was fixed by 0.1 M sodium cacodylate (pH 7.35) including 2% paraformaldehyde and 2.5% glutaraldehyde. The fixed cell pellet was mixed with 3% agarose solution, fixed in unbuffered aqueous solution containing osmium tetroxide, stained by 3% uranyl acetate solution, and the sample was rinsed 3 times by H₂O between every step. Sample was dehydrated using a graded series of diluted ethanol in H₂O. Then, sample was embedded in resin using Embed 812 kit and sectioned into 70 nm-thick blocks using a ultramicrotome (Leica Ultracut UC6, Leica Biosystems, Wetzlar, Germany). The thin section was stained uranyl acetate and lead citrate solution.

Cellular Ice formation analysis

TrampC1 cells (2 × 10⁵ cells) were seeded in a 35 mm diameter culture dish for overnight. When they fully adhered to the substrate, we replaced the growth media with control, E-GNPs, DF-INNs-containing growth media (0.2 mg mL⁻¹). 4 h after replacement, each culture dish was washed with PBS twice. Then, the dish was fixed in a cooling system on optical microscope (Olympus CKX41) equipped with digital camera (AmScope MU 1000, CA, USA). Then, cryochamber cooled down the system using −80°C cryogen and OM images were taken every 30 s.

Animal

All experimental procedures were approved and reviewed by the Institutional Animal Care and Use Committee of Northwestern University (IS00024599). C57BL/6 mice (male, 8–10 weeks old, 18–30 g) (Charles River Laboratories, MA, USA) were utilized in this study. Mice were anesthetized with 2% isoflurane (Isothesia, Abbot Laboratories, IL, USA) with 1 L min⁻¹ oxygen.

Anti-tumor effects in TrampC1-bearing model

TrampC1 cells (2 × 10⁶ cells mice⁻¹, in 100 μL of saline) were subcutaneously injected into the right flank of C57BL/6 mice. After three weeks, tumor bearing mice with around 75 mm³ volume were randomly divided into 3 groups as following: control, E-GNPs with cryotherapy (E-GNP+Cryo), and DF-INNs with cryotherapy (DF-INN + Cryo). Either E-GNPs or DF-INNs (10 mg mL⁻¹, 40 μL) were intratumorally injected (i.t.). Each sample was slowly injected (1 μL s⁻¹) to the center of tumor using the insulin syringe with an image-guided approach. A LN₂-based cryoprobe was inserted into the center of tumor 4 h after sample injection. Then, each mice had a cryotherapy composed of two consecutive freeze-thaw cycle: 1^st freezing (3 min), 1^st thawing (passive, 4 min), 2^nd freezing (3 min), and 2^nd thawing (passive, 4 min). During the process, mice were anesthetized with 2% isoflurane with 1 L min⁻¹ oxygen. Then, they were located on an electric heating pad to prevent hypothermia, and their body temperature were monitored using K-type thermocouple connected to a digital thermometer (Proster) on tumor, and thermal camera (ICI 7320, Infrared Cameras Inc. TX, USA). Tumor volume (V) was monitored using a caliper in daily for 14 days and calculated with following equation: V (mm³) = 0.5×(width)²×(length) (width ≤ length).

In vivo cryo-responsive intratumoral distribution of DF-INNs

X-ray CT studies were investigated using a micro-CT imaging system (NanoScan PET/CT, Mediso, MA, USA) at the Center for Translational Imaging at Northwestern University. For in vitro phantom imaging, concentration-dependent CT contrast of DF-INNs was investigated under 35 kVp of X-ray, and Hounsfield unit of each sample were measured with RadiAnt DICOM viewer (version 2022.1, Medixant, Poland). For the in vivo study, TrampC1 tumor syngeneic mice were imaged. Before imaging, tumor-bearing mice were anesthetized with 2% isoflurane with 1 L min⁻¹ oxygen. 40 μL of DF-INNs (20 mg mL⁻¹ in saline) was injected into a tumor center with CT guidance. The solution was injected with 1 mL insulin syringe with 29G needle, injection rate of 1 μL s⁻¹. Imaging parameters were the following: slice thickness, 103.25 μm; tube voltage, 35 kV; tube current, 1 mA; exposure time, 240 ms. Baseline images were acquired pre-injection and immediately after injection (day 0). Follow-up imaging was taken on post-treatment days 1, 3, 7. (n = 4 mice group⁻¹). Acquired DICOM images were analyzed and reconstructed into 3D model using 3D Slicer (v5.2.1.). The employment of water molecules in subcutaneous space 1 day after cryotherapy was confirmed by T₂-weighted MRI using Bruker 7.0 T ClinScan MRI (Bruker Biospin, Ettlingen, Germany).

Histology and Immunohistochemistry

To assess the distribution of DF-INNs, tumors were harvested from the mice or rats and fixed for 24 h in 4% paraformaldehyde. The tissue sections (5 μm) were stained with H&E, TUNEL, and Ki-67 (Promega, WI, USA). All slides were scanned using Nanozoomer 2.0-HT slide scanner (Hamamatsu, Shizuoka, Japan).

Immune cell characterization using flow cytometry

After sacrifice, tumor tissues cells were acquired, homogenized and sieved to exclude the extra tissue samples using a cell strainer (40 μm, Corning, NY, USA). Erythrocytes were lysed by using ACK lysis buffer (Thermofisher, MA, USA) for 5 min under room temperature. The number of purified cells were counted by Countess I (Thermofisher, MA, USA), and 1 × 10⁶ cells were stained with indicated antibodies: BV605 mouse anti-mouse CD45.1 (1:100; Lot. 2182084; Cat. 747743; Clone A20), FITC rat anti-mouse CD3 (1:100; Lot. B289551; Cat. 555274; Clone 17A2), V450 rat anti-mouse CD8a (1:100; Cat. 560471; Clone 53–6.7), and BV421 rat anti-mouse CD274 (1:100; Lot. 1222292; Cat. 564716; Clone MIH5) (BD Biosciences, NJ, USA), to evaluate the immune cell distribution.

Statistical analysis and reproducibility

Data were analyzed using Origin 2017 (OriginLab, MA, USA) and Prism software version 5.0.0 (GraphPad, CA, USA). Graphical data are presented as means ± standard deviation. Statistical significance of differences between two groups was determined using Student’s unpaired t-test or one-way analysis of variance. P-values were considered statistically significant following stars: *p<0.05, **p<0.01, and ***p<0.001. Each experiment was independently repeated at least three times, including representative TEM, OM, confocal images.

Appendix A. Supporting Information

Supplemental Information is available from the online version at doi:

Data and materials availability materials

The raw and processed data required to reproduce these findings will be available upon request.