Bioinspired 3D hydrogel scaffold to mimic tumor microenvironment for investigating into the anoikis resistance mechanisms in colorectal cancer

Abstract

Anoikis resistance constitutes a critical pathophysiological mechanism driving metastatic progression in colorectal cancer. While in vitro models are essential for mechanistic studies, conventional 2D cultures inadequately replicate tumor microenvironment (TME) complexity. In this study, we developed a biomimetic composite hydrogel (GHP_4a) composed of 4-arm-PEGDA, Gelatin Methacryloyl, and Hyaluronic acid Methacryloyl to establish compact 3D cell spheres that mimic TME. The GHP_4a demonstrates superior biocompatibility, suitable mechanical strength at 600–700 Pa, and biomimetic properties that promote cell proliferation and differentiation when co-culturing Caco-2 cells. The cell anchorage and survival for anoikis resistance are enhanced compared to traditional 2D cell incubation due to a well-organized 3D formation akin to the extracellular matrix (ECM). The biomimetic mechanism may be attributed to the fact that the GHP_4a promoted the activation of key pro-survival pathways, including FAK and PI3K/Akt signaling, and suppressed caspase-mediated apoptosis. Single-cell RNA sequencing revealed distinct transcriptional profiles within the proliferating T cell population, suggesting a novel regulatory mechanism of T cell-mediated anti-tumor immunity in the TME. Additionally, radiomic analysis identified significant differences in tumor heterogeneity and texture between GHP_4a-based 3D cultures and traditional 2D models. These findings established the GHP_4a as an effective candidate for the in vitro tumor anoikis resistance model, providing a unique approach for studying anoikis resistance in colorectal cancer and offering a robust tool for the development of advancing cancer diagnosis and therapy strategies.

Graphical abstract

1. Introduction

Colorectal cancer is one of the most prevalent malignancies globally and ranks fourth among the leading causes of cancer-related mortality. By 2040, its global incidence is projected to rise by 45 % to 1.8 million cases annually, while related deaths may increase by 50.4 % to 1.3 million [1,2]. The rising incidence and mortality of colorectal cancer pose significant challenges to clinical treatment [3,4]. However, current standard therapies, e.g., surgery, often fail to significantly improve survival. These clinical difficulties primarily stem from the cancer's high malignancy and inherent resistance to treatment [5]. Thus, identifying novel prognostic markers and therapeutic targets is essential for improving treatment strategies. Nevertheless, the highly malignant and complex pathology of colorectal cancer [[6], [7], [8]] severely hinders the direct in vivo detection and extraction of pathological biomarkers. Developing an in vitro model that accurately mimics the progression of colorectal cancer could enable more rapid and sensitive detection of cancer biomarkers, thereby enhancing early diagnosis and therapeutic intervention.

In recent years, anoikis resistance has been recognized as a hallmark of tumor metastasis [9]. Anoikis is a specialized form of programmed cell death triggered when anchorage-dependent cells detach from the ECM, leading to apoptosis. Anoikis resistance is a key feature in colorectal cancer progression and represents a potential therapeutic target for preventing metastasis. Under pathological conditions, colorectal cancer cells often develop mechanisms to evade anoikis-termed anoikis resistance-allowing them to survive in circulation and form metastatic colonies at distant organs [10]. Foundational studies highlighted the dynamic regulation of cytoskeletal proteins and signaling molecules, emphasizing the central roles of FAK (focal adhesion kinase) and PI3K/AKT pathways in anoikis resistance [11]. However, the mechanisms underlying anoikis resistance in colorectal cancer remain to be explored in greater depth. Meanwhile, developing effective in vitro models to investigate the anoikis resistance remains challenging because accurately replicating the complexity of the colorectal cancer microenvironment is inherently difficult [12].

Owing to their ECM-mimicking capabilities, including hydrophilic triple-network structure, excellent biocompatibility, and adjustable swelling properties, naturally derived hydrogels have been extensively applied in developing in vitro cancer models [[13], [14], [15]] and serve as effective tools to investigate the mechanisms of anoikis resistance. Cui et al. demonstrated the potential of hydrogels in tumor studies by using ECM-mimicking hydrogels to recreate realistic cellular microenvironments [16]. Furthermore, Gai et al. employed hydrogels as ECM substitutes in organoid cultures, offering a robust platform for disease-specific model construction [17]. Although several in vitro studies have explored anoikis resistance [18,19], constructing functional microenvironment models that accurately simulate apoptosis resistance remains a significant challenge. Besides, three-dimensional (3D) models built by hydrogel can more accurately replicate these complex dynamics than traditional two-dimensional (2D) cultures, providing deeper insight into tumor progression and supporting therapeutic development [20,21]. Therefore, developing a 3D in vitro colorectal cancer model based on hydrogel scaffold that simulates the TME of anoikis resistance is essential for elucidating metastatic mechanisms and guiding the discovery of targeted therapeutic strategies.

In this study, we developed a novel biomimetic polymer composite hydrogel-GHP_4a-that mimics the ECM (Scheme 1) and serves as a qualified 3D scaffold for spheroid culture, providing a robust in vitro platform for investigating the anoikis mechanism (Fig. 1). Specifically, GHP_4a, characterized by strong biomimetic features and high compatibility with microenvironments, was synthesized via a self-assembly strategy integrating Gelatin Methacryloyl (GelMA), Hyaluronic acid Methacryloyl (HAMA), and 4-arm-PEGDA. GelMA is synthesized by modifying gelatin with methacrylate anhydride, replacing free amine groups while preserving the arginine-glycine-aspartic acid (RGD) sequence essential for cell adhesion, which can form gel in a fast time [22]. HAMA then forms a more stable hydrogel network via photopolymerization and retains the biological activity of hyaluronic acid (HA), a key ECM component involved in cell adhesion, proliferation, and migration. While the flexible 4-arm-PEGDA reinforces the hydrogel network, providing toughness and enabling the formation of multiple hydrated layers to encapsulate bioactive cargo. Its physicochemical properties and in vivo applicability were subsequently evaluated. Comparative in vitro experiments assessed tumor growth and anoikis using 3D Caco-2 cell spheroids cultured on GHP_4a scaffold, ultra-low attachment plates, and conventional 2D substrates. The composite hydrogel GHP_4a containing these functional components exhibits improved structural stability structure and adaptability in vivo, positioning it as a promising ECM substitute that mimics the native TME, thereby enhancing cell viability and function and improving tissue engineering efficiency. Differential anoikis responses were further investigated in vivo using tumor-bearing nude mice. Bioinformatic analyses and CT imaging tests were employed to elucidate the molecular mechanisms underlying metastasis. This study provides valuable insights into engineering in vitro TMEs for efficient tumor biomarker capture.

Fig. 1.

The schematic diagram illustrates the preparation of 3D hydrogel as a composite of three materials for replacing ECM to construct an Anoikis resistance model to reduce cancer cell metastasis.

2. Materials and methods

2.1. Materials

Type B gelatin was purchased from Yuanye Biological Co. (Shanghai, China), and Sodium hyaluronate was purchased from Bloom Age Biotechnology Co. Ltd. (Shandong, China). Methacrylic anhydride was obtained from Sigma-Aldrich, St. (Louis, USA), PEGDA was purchased from EFL Company (Suzhou, China), 4-arm-PEG Acrylate was purchased from Ponsure Biotechnology Co. (Shanghai, China), and Lithium phenyl (2,4,6-trimethyl benzoyl) phosphinate (LAP) was purchased from Zesheng Technology Co. (Anhui, China). All other chemicals and solvents are of analytical grade.

2.2. Synthesis of hyaluronic acid methacryloyl (HAMA)

HAMA was synthesized as described previously [23]. Specifically, 2 g HA was dissolved in 300 mL aqueous solution at room temperature for complete dissolution. Then, 15 mL of methacrylic anhydride was slowly added to the solution using a constant pressure funnel, and the pH of the mixture during the reaction was maintained at 8–10 by adding sodium hydroxide. After continuous reaction at 4 °C ice bath for 12 h, the resulting conjugates were dialyzed in H₂O for 5 days. The product was obtained by lyophilization.

2.3. Synthesis of Gelatin Methacryloyl (GelMA)

GelMA was synthesized as previously described [24]. Briefly, 6 g type B gelatin was dissolved in 150 mL ultrapure water at 40 °C. Then 4.8 mL of methacrylic anhydride was added to the solution drop by drop, and the reaction proceeded for 2 h at 40 °C. Then, the resulting solution was dialyzed in H₂O for 3 days. The product was obtained by lyophilization.

2.4. Hydrogel preparation

To form GelMA-HAMA-4-arm-PEGDA (GHP_4a) hydrogels, various contents of GelMA, 4-arm-PEGDA, and 0.5 % HAMA mixture were dissolved in appropriate amounts of PBS containing 0.25 % photo-initiator LAP. The precursor solution was placed in Polytetrafluoroethylene (PTFE) models and photo-crosslinked by a light-emitting diode (LED) light (405 nm, 3 W). The obtained hydrogel was named GHP_4ax-y, where x and y represented the initial polymer concentration of GelMA and 4-arm-PEGDA, respectively.

2.5. Physicochemical characterizations

2.5.1. Nuclear magnetic resonance (NMR)

The chemical structure of HA, HAMA, gelatin, GelMA, 4-arm-PEGDA (Mn = 2000 Da), and GHP_4a hydrogels was characterized by ¹H NMR spectra on an NMR spectrometer (Ascend 400, 400 MHz, Bruker) using D₂O as the solvent, and the spectrum was recorded to confirm the degree of modification.

2.5.2. Fourier transform infrared (FTIR)

FTIR spectrometry (Nicolet iS50, ThermoFisher Scientific) was used to characterize the formation of GelMA, HAMA, 4-arm-PEGDA (Mn = 2000 Da), and GHP_4a.

2.5.3. Rheological properties

Hydrogel samples were prepared in cylindrical PTFE models (10 mm diameter, 9 mm height) for rheological properties. The rheometer (DHR-1, USA) was equipped with a temperature-controlled Peltier plate system with a gap size of 1 mm. For the Time-sweeps, the strain was set at 1.0 %. The frequency was set at 1.0 Hz, which simulated the normal physiological stride frequency. All measurements were performed at 37 °C. The Frequency-sweeps were carried out on crosslinked hydrogels at a strain of 1 % and covered the frequency range from 0.1 to 10 Hz. The Strain-sweeps ranging from 0.1 % to 10 % at 1 Hz frequency were performed to examine dynamic yielding properties, and the storage modulus (G′) and loss modulus (G″) were monitored.

2.5.4. Swelling property

Hydrogel samples were prepared in cylindrical PTFE models (10 mm diameter, 9 mm height) for the swelling test. Immersing the hydrogels in PBS at 37 °C and weighing at the predetermined time to monitor the increased weight. The swelling degree (SD) of samples was calculated as follows, as previously described [25]:

$$SD=\frac{W_s-W_i}{W_i}\times 100\%$$

where $W_s$ was the sample weight at different time points, and $W_i$ was the initial weight of hydrogels.

This test was conducted in triplicate, and average values were calculated.

2.5.5. Biodegradability-stability performance

Hydrogel samples were prepared in cylindrical PTFE models (10 mm diameter, 9 mm height) for degradation evaluation. Initial weights ($W_i$) of the hydrogels were determined after lyophilization. The samples were incubated in PBS solution at 37 °C, 200r/min for 48 h. At predetermined time points (14 days), the samples were rinsed with ultrapure water, freeze-dried, and weighed again ($W_t$). The percent remaining weight ratio (RW) was determined according to the following equation [26]:

$$RW=\frac{W_t}{W_i}\times 100\%$$

This test was conducted in triplicate, and average values were calculated.

2.5.6. SEM

Scanning electron microscopy (SEM, JEOL, JSM-IT800, Japan) was utilized to study the micromorphology.

2.5.7. Porosity

Hydrogel samples were prepared in cylindrical PTFE models (10 mm diameter, 9 mm height) for the porosity property. Initial weights ($W_i$) of the hydrogels were determined after lyophilization. Then, the samples were in anhydrous ethanol for 2 h, removed, and weighed again ($W_d$). The density of ethanol is ρ, and the volume of hydrogel is V. The porosity was calculated as follows [27]:

$$Porosity=\frac{\left(W_d-W_i\right)V}{\mathrm{\rho }}$$

This test was conducted in triplicate, and average values were calculated.

2.5.8. Mechanical properties

Hydrogel samples were prepared in cylindrical PTFE models (10 mm diameter, 9 mm height) for compression test. Specifically, the loaded samples were both in the wet state and measured at a compression speed of 1 mm min⁻¹ using a Discovery DHR-1 rheometer. The maximum strain was determined from the peak of the stress-strain curve [28].

2.6. Biocompatibility evaluation in vitro

Human colon adenocarcinoma-derived cell lines Caco-2 were obtained from the American Type Culture Collection (ATCC; cat. nos). 1640 complete medium was purchased from Fuheng Biology Co. (Shanghai, China), trypsin-EDTA was purchased from Thermo Fisher Scientific (Shanghai, China), phosphate-buffered saline (PBS), Calcein, AM and PI, Cell Counting Kit-8 (CCK-8) were purchased from KeyGEN BioTECH Co., Ltd. (Nanjing, China). All chemical substances were employed straight away without any additional purification.

2.6.1. Cell culture

Caco-2 cells were cultured in 1640 complete medium on 100 mm cell culture plates at 37 °C with 5 % CO_2. The 1640 complete medium comprised 10 % Fetal bovine serum (FBS), Penicillin, and Streptomycin. The cells were passaged and used for the cytocompatibility tests at 80 % confluency by trypsin/EDTA solution. For 3D cell culture, Caco-2 cells were first resuspended in hydrogels with a density of 1 × 10⁶ cells/mL and photoclicked by a 405 nm light. The encapsulated hydrogels were washed with PBS three times and cultured with 1640 complete medium.

2.6.2. Cell activity (live/dead) assay

To directly observe the viability of cells, the samples were stained using a live/dead assay kit following the manufacturer's protocol. The stained samples were observed using a fluorescence microscope (Nikon Ti2, Japan) or a laser scanning confocal microscope (Nikon TE2000, Japan), and images of live and dead cells stained green and red were captured.

2.6.3. Cell adhesion and proliferation- CCK-8

To assess the adhesion ability of the cells cultured on the hydrogels, the cells were seeded on the sterilized hydrogel in a 96-well plate and incubated for 2, 4, and 6 days. At each time point, a live/dead assay was conducted, and CCK-8 was used to analyze cell proliferation on the surface of hydrogels. 20 μL CCK-8 and 100 mL complete media were added to each well, and the cells were incubated in the dark for 4 h at 37 °C. After incubation, culture media were transferred to a new 96-well plate, and the absorbance at 450 nm was immediately recorded. Three samples were tested for each group.

2.6.4. Forming efficiency assay

To evaluate the accumulation and imaging of EcN and CMCB in tumors, the mice bearing tumors (Caco-2) at a size of ∼100 mm^3, as described in the section of subcutaneous tumor models, were intravenously injected with 1 × 10⁷ CFUs of EcN or CMCB. All the mice were imaged by an in vivo imaging system and sacrificed at predetermined time points.

2.6.5. Western blot analysis

Total proteins of the Caco-2 cells (2D, GHP_4a2-5, and GHP_4a3-5) were isolated on ice with radioimmunoprecipitation assay (RIPA) lysis buffer, and the protein concentration was measured using a BCA protein assay kit (Beyotime, China). A total of 40 μg of proteins were separated by SDS-PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane. The PVDF membranes were blocked with 1 % Bovine Serum Albumin (BSA) for 1 h and then incubated with the primary antibody at 4 °C overnight. The next day, the PVDF membranes were incubated with a secondary antibody for 1 h at room temperature. Washing with Western Washing Buffer (Beyotime, China) 3 times, the PVDF membranes were further developed with the ECL Plus Western Blotting Detection Reagents, and the signals were detected using the ChemiDoc Imaging System (Bio-Rad, USA). The antibodies used in this study are MCL-1(Aifang Biological, China) and HRP-conjugated AffiniPure Goat Anti-Rabbit IgG H&L (MedChemExpress, USA).

2.6.6. Drug treatment

Drug sensitivity testing was performed using 50 × 10⁻³ mM 5-FU as a model drug. For these conditions, Caco-2 cells were cultured in a 96-well plate, corresponding to a density of 2 × 10⁶ cells/well, then cells were treated with rifampicin in culture media for 48 h.

2.6.7. In vivo antitumor experiment

Female BALB/C nude mice (4–5 weeks old, 12–16g, 21 mice) were bilaterally injected subcutaneously with 5 million Caco-2 cells (2D and cell embedded in the GHP_4a2-5 and GHP_4a3-5) diluted with 100 μL phosphate-buffered saline containing 50 % PBS. Once palpable tumors are established, the tumor volume of the mice will be recorded. Tumor weight was measured at the time of killing. All animals were measured by two investigators who were blinded to group allocation. Animal experiments were performed following the university laboratory animal guidelines and with approval from the Animal Experimentation Ethics Committee of the Second Affiliated Hospital of Soochow University.

2.6.8. Histological biochemical analysis

Animal management procedures were approved by the Animal Ethics Committee of Soochow University (number: 202312A0412). Tumor sections were incubated with the antibodies Ki-67 (1:200), Bad (1:200), Mcl-1 (1:200), PARP-1 (1:100), and Survivin (1:200) overnight at 4 °C. The remaining steps were performed using the DAKO kit (Aifang Biotechnology, Hunan, China). The antibodies used are derived from Aifang Biotechnology. Quantitative analysis of H&E and immunohistochemistry (IHC) images was performed by ImageJ software (NIH, USA; v1.53), with a standardized process consisting of the following steps:

Preprocessing: the original RGB images were converted to 8-bit grayscale maps, and the background was eliminated by optical density calibration (OD = log₁₀ (255/grayscale value)). Interference, and the color deconvolution plug-in was used to separate the DAB color channels for double-stained samples; Signal segmentation: positive areas were defined by thresholding (manual adjustment or automatic classification by the IHC Profiler plug-in), and necrotic areas and nonspecific staining were excluded; Core quantification: the area of the target area (μm²) was measured with the integrated optical density value (IOD), and the average optical density (AOD = IOD/Area) was calculated and percentage of positive cells; Quality control validation: batch processing using IHC Toolbox plug-in ensured parameter consistency, 20 % of randomly selected samples were double-blind reviewed by two pathologists (Interclass Correlation Coefficient (ICC) > 0.85), and the algorithm agreed with the manual scoring by 88.6 % [29]; Statistical analysis: Data were exported to GraphPad Prism 9.0 with ANOVA or t-test for continuous variables and χ² test for categorical variables (P < 0.05). Screenshots of the full process parameters were archived, and experiments were done in a College of American Pathologists (CAP)-certified laboratory.

2.6.9. RNA sequencing

At 35 days post-modeling, the tumor tissue inoculated subcutaneously in nude mice was dissected and obtained. Tumor tissues were placed in cryogenic bottles, labeled according to their respective groups, rapidly frozen in liquid nitrogen for 2 min, and subsequently stored at −80 °C until sequencing was required. Samples were submitted to GENEWIZ for transcriptome sequencing and analysis. Inc. Suzhou.

2.6.10. Data progressing and dimensionality reduction

Raw count data from RNA sequencing were processed and analyzed using R (version 4.2.0). Uniform Manifold Approximation and Projection (UMAP) was performed using the 'umap' package for dimensionality reduction and visualization. Prior to UMAP analysis, count data was normalized using variance stabilizing transformation (VST) implemented in DESeq2 to account for variance-mean dependency in count data. The UMAP visualization was generated to compare expression patterns across three experimental groups: GHP_4a2-5, GHP_4a3-5, and control 2D, with three biological replicates per group.

2.6.11. Differential expression analysis

Differential gene expression analysis was conducted using the DESeq2 package (version 1.34.0). We performed two main comparisons: GHP_4a2-5 vs. 2D control and GHP_4a2-5 vs. GHP_4a3-5. Genes with adjusted p-values <0.05 and absolute log2 fold change >1 were considered significantly differentially expressed. Volcano plots were generated using the ggplot2 package to visualize the distribution of differentially expressed genes, with red points indicating significantly regulated genes and grey points representing non-significant changes. For visualization of key apoptosis-related gene expression patterns, we constructed heatmaps using the pheatmap package. Expression values were row-wise standardized (z-score), and hierarchical clustering was applied to both rows and columns using Euclidean distance with the complete linkage method. To evaluate cellular composition changes across different treatment conditions, we employed the CIBERSORTx algorithm for cell fraction analysis. Corresponding single-cell data in.h5 format and annotation results were downloaded from the TISCH database. The single-cell data were processed and analyzed using MAESTRO (version 1.5.1) and Seurat (version 4.1.0) packages. Following a standard workflow of quality control, normalization, and batch effect correction, cells were re-clustered using the t-SNE method.

2.6.12. Functional enrichment analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using the clusterProfiler package (version 4.2.2) to identify biological functions associated with differentially expressed genes (DEGs). Two separate analyses were conducted for DEGs identified in GHP_4a2-5 vs. control 2D and GHP_4a2-5 vs. GHP_4a3-5 comparisons. Gene symbols were converted to ENTREZ IDs using the org.Hs.eg.db annotation package. For GO enrichment analysis, we examined three categories: biological process (BP), cellular component (CC), and molecular function (MF), with particular focus on terms related to extracellular matrix components, membrane transport processes, and lipid metabolism. The KEGG pathway analysis was conducted using the human pathway database with a significance threshold of p < 0.05. The top enriched terms were visualized using ggplot2, with the dot size representing gene count and color intensity indicating statistical significance.

2.6.13. Protein-protein interaction network analysis

To investigate the potential functional interactions among differentially expressed genes (DEGs), we constructed a protein-protein interaction (PPI) network using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (version 12.0, https://string-db.org/). The identified DEGs were input into the STRING database to retrieve the protein-protein interactions. The minimum required interaction score was set to medium confidence (0.400), and disconnected nodes were hidden from the network. The resulting PPI network was visualized with nodes representing proteins and edges indicating functional associations between proteins.

2.6.14. Micro-CT imaging and radiomics analysis

Computed tomography (CT) scans were performed on mice from three treatment groups (GHP_4a2-5, GHP_4a3-5, and control 2D) using CT scanners. 3D tumor segmentation was conducted using 3D software to delineate tumor lesions. The segmentation process was performed semi-automatically with manual verification by an experienced radiologist to ensure accuracy.

Following tumor segmentation, quantitative imaging features were extracted from the regions of interest (ROIs). The extracted radiomics features included.

• •First-order statistics features
• •Shape-based features
• •Texture features (including GLCM, GLRLM, GLSZM)
• •Wavelet-transformed features

To visualize the differences in radiomics features among the three groups, a hierarchical clustering heatmap was generated using R with the 'pheatmap' package. The heatmap displayed the standardized feature values, with rows representing individual radiomics features and columns representing samples from different groups. Hierarchical clustering was performed using Euclidean distance and complete linkage method.

Additionally, bar plots were created to compare specific features of interest among the three groups using GraphPad Prism. Statistical significance was assessed using one-way ANOVA followed by Tukey's post-hoc test for multiple comparisons. Error bars represent standard deviation (SD), and p-values less than 0.05 were considered statistically significant.

2.7. Statistical analysis

Except as noted, all data were presented as mean ± standard deviation and were examined with GraphPad Prism 10.0. The error bar represents the standard error obtained from three separate measurements. The unpaired Student's bilateral t-test was used to assess the significance of the data: ∗P. (∗<0.05, ∗∗<0.01, and ∗∗∗<0.001)

3. Results and discussion

3.1. Physical-chemical characterizations of GelMA-HAMA-4-arm-PEGDA hydrogel

The GHP_4a hydrogel was synthesized via a photo-crosslinked process by integrating HAMA and 4-arm-PEGDA into the GelMA network. The successful modification was verified by ¹H NMR spectra of GelMA (Fig. S1) and HAMA (Fig. S2), which exhibited new peaks at 5.4–6.4 ppm corresponding to vinyl groups. In the ¹H NMR spectra of 4-arm-PEGDA (Fig. S3), new peaks between 5.8 and 6.5 ppm indicated the formation of acrylate double bonds, and a peak at approximately 4.3 ppm suggested the linkage between acrylate and PEG chains (Fig. S4). FTIR spectra (Fig. S5) further showed characteristic peaks between 3000 and 3500 cm⁻¹, supporting the successful bonding of GelMA, HAMA, and 4-arm-PEGDA.

Subsequent morphology analysis by scanning electron microscopy (SEM) revealed that the GHP_4a hydrogels exhibited a honeycomb-like porous architecture after freeze-drying (Fig. 2A and B), providing a suitable environment for cell seeding and adhesion. Interestingly, pore size decreased with the increase of 4-arm-PEGDA concentration. Specifically, the pore sizes were approximately 80 μm for GHP_4a2-3 (GelMA: 4-arm-PEGDA = 2:3 %) and 68 μm for GHP_4a2-5 (GelMA: 4-arm-PEGDA = 2:5 %). Hydrogels containing higher 4-arm-PEGDA concentrations showed a more regular and compact structure, whereas lower concentrations led to a more heterogeneous and loose morphology. Porosity analysis (Fig. 2C) indicated that although GHP_4a2-3 hydrogels exhibited the highest porosity, GHP_4a2-5 and GHP_4a3-5(GelMA: 4-arm-PEGDA = 3:5 %) showed a more uniform pore distribution. These results suggest that GHP_4a2-5 hydrogels possess optimal microporous structures for 3D cell culture.

Fig. 2.

Analysis and evaluation of the synthesized Gelatin Methacryloyl (GelMA)/Hyaluronic acid Methacryloyl (HAMA)/4-arm-PEGDA (GHP_4a) hydrogel Scaffold. (A) SEM (scanning electron microscopy) images. (B) Pore size statistics of each group of hydrogels. (C) Porosity of each group of hydrogels. (D) Compressive stress-strain curve. (E) Compressive strength (n = 5, ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001). (F) Compressive modulus (n = 5, ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001). (G) Swelling rate. (H) Degradation curve. (I) Time-sweep of hydrogels.

The mechanical properties of GHP_4a hydrogels were evaluated by compressive testing. GHP_4a2-5 hydrogels demonstrated greater strain-at-break and maintained their integrity without rupture under 75 % compression (Fig. 2D). In contract, GHP_4a2-1 (GelMA: 4-arm-PEGDA = 2:1 %), GHP_4a2-3, and GHP_4a3-5 hydrogels fractured at 74 %, 76 %, and 70 % strain, respectively. Increasing 4-arm-PEGDA content enhanced both compressive strength and modulus, with GHP_4a2-5 showing approximately a 220 % increase compared to lower concentration groups (Fig. 2E and F).

Hydrogels with various 4-arm-PEGDA concentrations were fabricated using 2000 Da 4-arm-PEGDA chains and photo-cross-linkable methacrylic anhydride (MA) to validate these findings further. Gelation time measurements (Fig. S6) indicated that hydrogels with 3–5 % 4-arm-PEGDA concentrations achieved rapid gelation (<60 s), while GHP_4a2-1 at 1 % concentration showed no effective gelation (>120 s).

Swelling behavior was then assessed. All hydrogels exhibited rapid swelling within 2 h, reaching 350–900 %, and plateaued after 24 h. GHP_4a2-1 and GHP_4a2-3 demonstrated faster dissolution rates and higher equilibrium swelling ratios than GHP_4a2-5 and GHP_4a3-5 (Fig. 2G). These results indicate that the swelling behavior was inversely correlated with 4-arm-PEGDA content, suggesting that denser network structures restricted water uptake and enhanced hydrogel stability.

Degradation behavior was evaluated by monitoring mass loss over time. After 14 days in PBS, GHP_4a2-5 hydrogels retained approximately 88 % of their original mass, while GHP_4a2-1, GHP_4a2-3, and GHP_4a3-5 hydrogels degraded more rapidly, retaining less than 85 % (Fig. 2H). These findings imply that GHP_4a2-5 hydrogels exhibit improved in vitro stability, likely due to their higher crosslinking density, making them more suitable for long-term tissue engineering applications.

The rheological data provide key experimental evidence for investing the mechanism in which matrix hardness of hydrogel formulations modulates cell attachment resistance. Previous studies have systematically elucidated the bidirectional regulatory properties of matrix stiffness: within the regular stiffness interval (>1 kPa), increased stiffness reduces adherence resistance by diminishing integrin-mediated stability of adhesion sites [30,31]; whereas within the ultrasoft interval (<1 kPa), decreased stiffness enhances adherence stability by expanding cell-matrix contact area, thereby strengthening attachment stability [32]. In this study, the mechanical properties of the selected formulations (GHP_4a2-1, GHP_4a2-3, GHP_4a2-5, GHP_4a3-5) were evaluated using rheological measurements. The results showed that all the hydrogels retained their structural integrity within the strain range of 0.1–1 % (Fig. S7A). This indicates that these hydrogels can withstand mechanical loading conditions typically encountered in cell culture (typical cell traction strain <5 %) without exhibiting significant creep or fracture. Consequently, they provide a robust matrix platform for conducting anoikis resistance experiments. Storage modulus (G′) and loss modulus (G″) exhibited concentration-dependent increases (Fig. 2I and Fig. S7B), with the average G′ values of 200, 450, 700, and 1000 Pa for GHP_4a2-1, GHP_4a2-3, GHP_4a2-5, and GHP_4a3-5, respectively. This gradient spans the range from super soft (<1 kPa) to medium hard (1–20 kPa), precisely aligning with the key regulatory domains. Among these samples, GHP_4a2-1 (200 Pa) and GHP_4a2-3 (450 Pa) fall within the ultra-soft interval, which can be employed to validate the “contact area expansion effect” proposed by Best et al. Meanwhile, GHP_4a2-5 (700 Pa) and GHP_4a3-5 (1000 Pa), located in the conventional interval, are appropriate for validating the “adhesion site stability model”. Notably, the G′ value of GHP_4a2-5 hydrogels was close to that of native intestinal tissue (600–700 Pa) [[33], [34], [35]], indicating matching biomimetic mechanical properties, highlighting the advantages of formulation design for modeling specific physiological environments. In addition, the G′ values are consistently higher than G'' (G' > G″), suggesting that the hydrogels are elastically dominated (tan δ = G''/G' < 1), which is associated with the energy dissipation mechanism of cellular adherence resistance. The rheological empirical evidence described above further substantiates the rationale for the formulation choice. Specifically, gradient G′ values (200–1000 Pa) are synchronized to capture the super-soft effect in contrast to the conventional model, thereby enabling a direct comparison of anoikis resistance thresholds across intervals; concentration-dependent G′ growth independent of the adherent ligand ensures an isolated contribution of hardness as a mechanical variable; furthermore, the bionic hardness of GHP_4a2-5 and GHP_4a3-5 are suitable for intestinal tissue studies, while the softer properties of GHP_4a2-5 effectively mimic the softening of tumor microenvironment. In summary, rheological measurements not only provide experimental validation for the modulation of matrix hardness, but also empirically link the theoretical hypotheses to the bionic matching and elastic properties via G′ values, thereby establishing a robust foundation for investigating into tissue engineering and disease modeling.

Overall, GHP_4a2-5 and GHP_4a3-5 hydrogels demonstrated favorable mechanical properties, swelling behavior, and degradation profiles, suggesting their potential as scaffolds for 3D intestinal cell culture in vitro and in vivo.

3.2. Biological characterizations of GelMA-HAMA-4-arm-PEGDA hydrogel in vitro

To evaluate the biocompatibility and cytotoxicity of the prepared hydrogels-key parameters for tissue scaffold engineering-the human colorectal cancer cell line Caco-2 was encapsulated in two types of GHP_4a hydrogels (GHP_4a2-5 and GHP_4a3-5) for 3D culture (Fig. S8).

Cell viability and proliferation were first assessed using the CCK-8 assay. The results indicated no significant difference in cell proliferation between GHP_4a2-5 and GHP_4a3-5 over a 7-day culture period. However, GHP_4a2-5 supported slightly higher proliferation on days 5 and 7, possibly due to its superior mechanical properties providing a more favorable environment for cell growth (Fig. 3A and B).

Fig. 3.

Biocompatibility of the synthesized Gelatin methacrylate (GelMA)/Hyaluronic acid methacrylate (HAMA)/4-arm-PEGDA (GHP_4a) hydrogel Scaffolds. (A) Scheme of 2D-3D culture model of colon cancer Caco-2 cell line. (A) Cell viability of 3D cultured Caco-2 cells cultured in hydrogels. (B) The proliferation of 3D cultured Caco-2 in hydrogels. (C) Drug treatment in different groups of hydrogels as-synthesized. (D) Live/dead cell ratio extracted from the staining fluorescent images (n = 5). (E) Live/dead staining fluorescent images of GHP_4a-Caco-2 hydrogel scaffolds on day 1, day 3, and day 5. (F) Images of Live/dead staining fluorescent changes of GHP_4a-Caco-2 hydrogel scaffolds on days 1, 3, and 5. (G) Survivin expression amount in the Caco-2 tumor extracted from western blot band. (I) Expression of Mcl-1 and Survivin in Caco-2 cells (2D, GHP_4a 2–5 and GHP_4a 3–5) by Western Blotting. Error bars indicate the statistically significant differences as ∗P Error bars indicate the statistically significant differences as ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Next, the drug responsiveness of these 3D models was evaluated. 5-Fluorouracil (5-FU) was administered to both GHP_4a2-5, GHP_4a3-5 3D cultures and a 2D monolayer culture. As shown in Fig. 3C, cell viability decreased in a drug concentration-dependent manner across all groups. At lower concentrations (15–60 μg/mL), the 3D models exhibited slightly reduced viability compared to the 2D model. In contrast, at higher concentrations (250–1000 μg/mL), the 3D cultures demonstrated greater sensitivity to 5-FU, reflected by a more pronounced decrease in cell viability. These differences may be attributed to slower drug diffusion and restricted exocytosis within the denser matrix of the 3D hydrogels. Initially, minimal differences were observed between 2D and 3D models, but over time, more efficient drug clearance in the 2D cultures contrasted with prolonged drug retention in the 3D matrix, leading to enhanced cytotoxicity.

Furthermore, live/dead cell staining revealed minimal red fluorescence in the Caco-2 cells encapsulated within GHP_4a2-5 and GHP_4a3-5 hydrogels, indicating excellent cell viability throughout the 7-day incubation period, with no significant differences compared to 2D cultures (Fig. 3D and E). 3D confocal microscopy imaging also demonstrated that most Caco-2 cells survived within the biocomposite hydrogels, maintaining spheroid morphology (Videos S1 and S2) without noticeable cell protrusion or spreading (Fig. 3F).

Supplementary data related to this article can be found online at https://doi.org/10.1016/j.mtbio.2025.102061

The following are the Supplementary data related to this article.

Western Blot analysis was performed to further explore the molecular basis of cell survival within the hydrogels and evaluate Mcl-1 protein expression in 2D, GHP_4a2-5, and GHP_4a3-5 cultured cells. Mcl-1 expression was significantly higher in GHP_4a2-5 cells compared to 2D cells (P = 0.0004), while no significant difference was observed between GHP_4a3-5 and 2D cells (P = 0.3886). A similar trend was observed for Survivin expression, with upregulation in GHP_4a2-5 and downregulation in GHP_4a3-5 relative to the 2D model (Fig. 3G, H, I, and Fig. S9). These findings suggest enhanced cell survival signaling within the GHP_4a2-5 matrix.

To assess whether GHP_4a hydrogels facilitate anoikis resistance, the morphology and growth behavior of Caco-2 cells were further investigated. Cells cultured within GHP_4a hydrogels formed suspended spheroids without substrate adhesion, and daily imaging confirmed continued cell viability and clusters formed over time [36]. Survival rates of approximately 90 % and 85 % were recorded for clusters formed within GHP_4a2-5 and GHP_4a3-5, respectively, suggesting that the GHP_4a hydrogels provided a supportive microenvironment with significant resistance to anoikis [37]. In general, conventional 2D culture systems, which often induce higher levels of anoikis due to limited cell-ECM interactions, are inadequate for establishing robust and quantifiable anoikis-resistant models in vitro [38]. These results indicate that 3D Caco-2 models based on GHP_4a hydrogels more closely replicate critical in vivo TME aspects, including enhanced drug response and survival signaling.

In conclusion, our findings demonstrate that GHP_4a hydrogel-based 3D scaffolds effectively promote Caco-2 cell survival, proliferation, drug responsiveness, and anoikis resistance. These 3D models offer a promising in vitro platform for tumor biology research and drug toxicity screening, providing a more physiologically relevant alternative to conventional 2D cultures.

3.3. In vivo assessment of GelMA-HAMA-4-arm-PEGDA scaffold in a nude mouse model

To evaluate the in vivo biological effect of hydrogel micro-spheroids, a subcutaneous tumor-bearing nude mouse model was established using 2D cultured and two types of 3D cultured cells (GHP_4a2-5 and GHP_4a3-5) and monitored for 35 days (Fig. 4A). Tumor volumes were measured every two days throughout the experiment, with growth curves shown in Fig. S10. During the experimental period, all groups exhibited comparable average weight gain rates, indicating good biosafety of the GHP_4a hydrogels without inducing adverse systemic effects (Fig. 4B).

Fig. 4.

In vivo anoikis resistance of GHP_4a. (A) Scheme of subcutaneous transplantation tumor model of colon cancer in nude mice. (B) Body weight and (C) tumor volume plots of different hydrogel groups during the period. (D) Tumor growth rate plots for different hydrogel groups were used during this time. (E) Photographs of the tumor after 34 days in different hydrogel groups and changes in tumor volume in mice of the GHP_4a2-5 group on the days 10, 15, 20, 25, 30, and the last day. Error bars indicate the statistically significant differences as ∗P, Error bars indicate the statistically significant differences as ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Notably, the tumor growth curves in Fig. 4C and D clearly illustrate the tumor development under the different treatments. All three groups exhibited similar growth rates during the initial phase (up to 15 post-implantation). However, tumors in the hydrogel groups subsequently demonstrated markedly accelerated growth, with the GHP_4a2-5 group showing a nearly two-fold increase in tumor volume compared to the 2D group. Specifically, the final tumor volumes in the GHP_4a2-5 group were approximately 1.5 times greater than those in the 2D group. These results suggest that the GHP_4a2-5 hydrogel provides a more favorable microenvironment for tumor proliferation in vivo. The enhanced tumor growth is likely attributed to improved cell survival and anoikis resistance [39], supported by the structural features of the GHP_4a scaffold that mimic the ECM. Finally, we terminated the model and excised the tumor tissues, and the photographs clearly illustrate the volume differences among the three models (Fig. 4E and Fig. S10).

To verify the presence of anoikis resistance in the tumor region, subcutaneously grafted tumor tissues were excised from the three groups of nude mice and subjected to HE staining and immunohistochemistry (IHC). As shown in Fig. 5A, anoikis-related proteins, including the tumor cell proliferation marker Ki-67 and anoikis resistance proteins Mcl-1 and Survivin, were highly expressed. In contrast, pro-anoikis proteins Bad and PARP-1 were expressed at low levels, consistent with the results in Fig. 3G, H, and I. Cells showed a better survival state in the two hydrogel groups compared to the 2D group, as reflected in the HE staining. The targeted proteins Ki-67, Mc1-1, and Survivin were more highly expressed in the GHP_4a2-5 and GHP_4a3-5 groups than in the 2D group.

Fig. 5.

In vivo anoikis resistance of GHP_4a. (A) H&E, Ki-67, Bad, Mc1-1, PARP-1, and Survivin tumor slice photographs after 34 days in different hydrogel groups. (B) Quantitative analysis results of HE, Bad, Ki-67, Mcl-1, PARP-1, and Survivin in nude mice in each group after 35 days of modeling. (C) H&E staining photographs of the main organs in different hydrogel groups, including the heart, liver, spleen, lung, and kidney. Error bars indicate the statistically significant differences as ∗P, Error bars indicate the statistically significant differences as ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

The targeted protein Bad expression was also lower in GHP_4a2-5 than in GHP_4a3-5 and group 2D. Notably, the quantitative analysis demonstrated that the GHP_4a2-5 group exhibited the highest expression of Ki-67, Mcl-1, and Survivin, alongside the lowest expression of Bad and PARP-1 among the groups (Fig. 5B). These results further validate the superior anoikis resistance behavior of the GHP_4a2-5 group. In addition, major organs, including the heart, liver, spleen, lung, and kidney tissues, were subjected to HE staining (Fig. 5C). No significant histopathological abnormalities were observed, further confirming the biosafety of the GHP_4a hydrogel used in vivo.

Potential molecular mechanisms underlying anoikis resistance in GHP_4a2-5 hydrogel were further proposed based on the observed tumor progression and protein expression patterns. Firstly, activation of the focal adhesion kinase (FAK) Pathway was likely triggered by RGD motif integrin interactions, promoting FAK phosphorylation, which stabilizes Mcl-1 expression and facilitates Bad inactivation, thereby preventing mitochondrial-mediated apoptosis. Secondly, downstream activation of the PI3K/Akt signaling pathway may further enhance survival and attenuate pro-apoptotic signaling. Thirdly, suppressed cleavage of PARP-1 and elevated Survivin expression suggest reduced activation of caspase-3 and caspase-9, contributing to apoptosis inhibition. Lastly, the dynamic ECM mimicry provided by the hydrogel likely sustained FAK and PI3K/Akt signaling through matrix remodeling, reinforcing cell survival. These findings confirm that 2D-cultured cells implanted in vivo are more prone to anoikis due to the lack of a supportive microenvironment.

In contrast, the GHP_4a2-5 composite hydrogel provides a biomimetic scaffold that promotes epithelial cell survival, enhances cell-matrix interactions, and alleviates anoikis. Finally, a comprehensive evaluation of five key indices related to anti-cell death mechanisms, including anoikis, ferroptosis, and cuproptosis [[40], [41], [42], [43], [44]], was performed (Fig. S11). The GHP_4a2-5 composite hydrogel performed superior to similar biomaterials, particularly in promoting anoikis resistance and simulating ECM functions.

In summary, the well-organized integration of GelMA, HAMA, and 4-arm-PEGDA into the GHP_4a hydrogel creates a biomimetic microenvironment [45] that enhances anoikis resistance in colorectal cancer cells, providing valuable insights into tumor progression and metastasis. The unique features of the GHP_4a hydrogel include [1]: A hydrated matrix resembling native hyaluronic acid, supporting cellular organization, enhancing cell-cell communication, and facilitating cell anchorage [[46], [47], [48]]. [2] RGD [22] motifs that activate integrin-mediated survival pathways (FAK/PI3K/Akt), which collectively reduce susceptibility to anoikis [49,50]. [3] Mechanical properties enhance mechanotransduction, preserve cytoskeletal integrity, and promote cell attachment [4]. Facilitation of cellular aggregation further strengthens survival signaling and anoikis resistance.

3.4. Differential expression of genes across distinct groups

To gain deeper insights into the molecular mechanisms by which the TME influences colorectal cancer cell growth and survival, we further explored how the biomimetic hydrogel scaffold GHP_4a affects anoikis resistance. Through its unique 3D structure and bioactive components, GHP_4a hydrogel may activate specific cellular signaling pathways, thereby altering gene expression profiles. To test this hypothesis, we conducted a comparative transcriptomic analysis between 2D cultured cells and GHP_4a microspheres.

Firstly, we performed transcriptome mRNA sequencing of 2D and GHP_4a microspheres (Fig. 6). Volcano plot analysis identified significant transcriptional changes between treatment groups. Specifically, 115 differentially expressed genes (DEGs) were identified between GHP_4a2-5 and control 2D, with 59 upregulated and 56 downregulated genes (Fig. 6A). Meanwhile, while 246 DEGs were found between GHP_4a2-5 and GHP_4a3-5, with 159 upregulated and 87 downregulated genes (Fig. 6B) (adjusted p-value <0.05, |log2FC| > 1). Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction analysis revealed distinct clustering patterns among the three treatment groups (GHP_4a2-5, GHP_4a3-5, and control 2D) (Fig. 6C). The precise spatial separation of samples in the UMAP plot demonstrated that the different treatment regimens induced distinct transcriptional profiles, suggesting treatment-specific effects on gene expression patterns. Hierarchical clustering of apoptosis-related genes, including Bad, Mcl-1, MKi-67, PARP-1, and Survivin, showed differential expression patterns across the treatment (GHP_4a) groups (Fig. 6D). Furthermore, cell fraction analysis revealed distinct cellular composition profiles among the treatment conditions, with variations in T cells, NK cells, B lineage, cytotoxic lymphocytes, fibroblasts, myeloid dendritic cells, CD8 T cells, endothelial cells, monocytic lineage, and neutrophils (Fig. 6E). These results indicate substantial transcriptional rewiring and cellular composition changes in response to different treatment approaches.

Fig. 6.

Transcriptomic analysis of different treatment groups. Volcano plots showing differentially expressed genes (DEGs) between GHP_4a2-5 and GHP_4a3-5 (A) with 59 upregulated and 56 downregulated genes and between GHP_4a2-5 and control. (B) with 159 upregulated and 87 downregulated genes (adjusted p-value <0.05, |log2FC| > 1). UMAP plot(C) reveals distinct clustering patterns among the treatment groups (GHP_4a2-5 in red, GHP_4a3-5 in yellow, and control in blue). Heatmap (D) shows differential expression of key apoptosis-related genes (BAD, MCL1, MKI67, PARP1, and SURVIVIN) across treatment groups. (E) Bar charts (bottom) display cell fraction analysis showing the distribution of different cell types across three replicates of each treatment condition.

3.5. Biological function enrichment of the DEGs

We conducted comprehensive pathway enrichment analyses to understand the functional implications of the transcriptional changes observed across treatment groups. GO enrichment analysis revealed that DEGs between GHP_4a2-5 and GHP_4a3-5 (Fig. 7A) were predominantly associated with ECM components and lipid metabolism, including high-density lipoprotein particles, plasma lipoprotein particles, basement membrane, blood microparticle, collagen-containing ECM, endoplasmic reticulum lumen, and glycosaminoglycan binding. The analysis also identified significant enrichment in biological processes related to coagulation regulation and homeostasis. In contrast, DEGs between GHP_4a2-5 and control (Fig. 7C) were significantly enriched in membrane transport processes, acute-phase response, serine-type endopeptidase inhibitor activity, and kidney development (p < 0.05), indicating distinct microenvironmental adaptations.

Fig. 7.

Functional enrichment analysis of differentially expressed genes. The left column shows an analysis of DEGs between GHP_4a2-5 and GHP_4a3-5, and the right column shows an analysis of DEGs between GHP_4a2-5 and control. (A, C) GO enrichment analysis shows significant biological processes and cellular components, with enrichment factor plotted on the x-axis and color/size indicating significance level and gene count. (B, D) KEGG pathway analysis highlighting significantly enriched pathways with gene ratio on the x-axis and dot color/size representing p-value and gene count. (E, F) Protein-protein interaction (PPI) network analysis reveals functional connections among DEGs, with nodes representing proteins and edges indicating interaction strength between proteins.

KEGG pathway analysis further differentiated the treatment effects. The GHP_4a2-5 versus control comparison (Fig. 7B) highlighted RNA transport, regulation of the actin cytoskeleton, lipoprotein metabolism, spliceosome, and cholesterol metabolism pathways. Meanwhile, GHP_4a2-5 versus GHP_4a3-5 comparison (Fig. 7D) revealed significant enrichment in complement and coagulation cascades, HIF-1 signaling pathway, glycolysis/gluconeogenesis, bile secretion, and insulin secretion (p < 0.05), suggesting metabolic reprogramming and hypoxia adaptation in the hydrogel environment.

Protein-protein interaction networks (Fig. 7E and F) identified key functional modules and hub genes mediating these responses. The networks for both comparisons revealed complex interconnections between differentially expressed genes, with distinct modules forming based on biological functions. Notably, the GHP_4a2-5 versus control network showed interconnected nodes involved in RNA processing and lipid metabolism. In contrast, the GHP_4a2-5 versus GHP_4a3-5 network displayed strong interactions between genes involved in complement cascades, metabolic pathways, and cellular survival. These findings demonstrate that GHP_4a2-5 hydrogel induces specific transcriptional programs that enhance cancer cell survival through coordinated regulation of ECM interactions, metabolic adaptation, and apoptosis resistance pathways.

3.6. Single-cell RNA sequencing analysis

Single-cell RNA sequencing analysis uncovered significant cellular heterogeneity within the TME, with distinct expression patterns of anoikis-related genes across immune populations. UMAP visualization (Fig. 8) revealed that proliferative T cells (Prolif T) exhibited a unique molecular signature characterized by the co-expression of both pro-apoptotic and survival factors. The pro-apoptotic regulator BAD showed moderate expression across multiple lineages with notable enrichment in Prolif T cells (∼0.6), while the proliferation marker MKI67 displayed highly restricted expression almost exclusively in this population (∼0.8), confirming their actively dividing state. Among anti-apoptotic factors, Survivin demonstrated remarkably selective expression in Prolif T cells (∼0.7) with minimal detection in other populations. At the same time, Mcl-1 (Fig. S12) exhibited a more heterogeneous pattern with the highest expression in CD8⁺ effector T cells (∼0.75), followed by monocytes/macrophages and Prolif T cells. PARP1, involved in DNA repair and apoptosis regulation, showed broad distribution across multiple immune subsets with particular enrichment in Prolif T and B cells.

Fig. 8.

(A–D) The t-SNE plot of the expression distribution of selected genes in different cells, where different colors represent expression abundance. The darker the color, the lower the expression of the gene in the cell, and the brighter the color, the higher the gene expression in the cell.

These findings provide granular insights into the molecular regulation of cell survival within distinct immune populations of the TME. The co-expression of both pro-survival factors and apoptotic regulators, specifically within proliferating T cells, suggests a delicate balance between expansion and vulnerability. This balance may be differentially affected by the hydrogel microenvironment, with important implications for understanding T cell persistence and function during immunotherapy approaches targeting the TME.

3.7. Radiomic analysis

Quantitative analysis of radiomics features revealed substantial differences among the GHP_4a2-5, GHP_4a3-5, and 2D groups, with particularly notable distinctions between the GHP_4a2-5 and GHP_4a3-5 groups (Fig. 9). The most significant variations were observed in texture and intensity-based features, with LargeAreaLowGrayLevelEmphasis showing the highest fold change (50.60-fold) in the GHP_4a2-5 group compared to controls, indicating more extensive areas of homogeneous low-intensity regions. The GHP_4a3-5 group demonstrated intermediate values, suggesting a distinct biological response pattern. LargeDependenceLowGrayLevelEmphasis (21.64-fold higher in GHP_4a2-5 vs. 2D) and contrast (11.27-fold increase in GHP_4a2-5 vs. GHP_4a3-5) revealed marked differences in tissue density patterns. Tumor heterogeneity markers showed group-specific patterns, with ZoneVariance (8.70-fold) notably elevated in the GHP_4a3-5 group, suggesting increased spatial complexity. At the same time, Busyness (5.81-fold) was predominantly higher in the GHP_4a2-5 group, indicating more rapid spatial changes in intensity patterns.

Fig. 9.

A. Radiomic pipeline of the study. B. The snapshot of 3D splicing software. C. Heat map of radiomic features. D. Comparison of radiomic features among three groups.

The analysis further revealed moderate to subtle variations in grey-level dependent features, with LongRunLowGrayLevelEmphasis (3.98-fold) particularly distinctive in the GHP_4a2-5 group, suggesting more extended regions of uniform intensity. The GHP_4a3-5 group showed intermediate values for JointEnergy (3.55-fold) and ClusterProminence (2.45-fold), positioning it between the GHP_4a2-5 and 2D groups regarding textural complexity. First-order statistical features, such as the 90th Percentile (2.40-fold), demonstrated a gradient effect across groups (GHP_4a2-5 > GHP_4a3-5 > 2D), reflecting systematic differences in overall intensity distributions. Shape-based features, while showing more modest variations, revealed that the GHP_4a2-5 group lesions tended toward greater flatness (1.49-fold) and lower sphericity (1.11-fold) compared to both GHP_4a3-5 and 2D groups, suggesting treatment-specific effects on tumor morphology. These comprehensive radiomics signatures distinguish between treatment groups and provide insights into the biological responses to different therapeutic approaches.

These quantitative imaging biomarkers collectively suggest that the GHP_4a2-5 treatment regimen induces the most pronounced textural changes, while the GHP_4a3-5 protocol results in intermediate alterations, both are significantly different from the 2D control group. The distinct patterns observed across feature categories indicate that each treatment protocol may affect tumor biology through various mechanisms, which could have important implications for treatment selection and monitoring.

Collectively, these multi-faceted analyses illustrate how the interplay between these pathways, mediated by the 3D GHP_4a hydrogel environment [[51], [52], [53], [54]], creates a protective niche for tumor cells in vivo. This model is prospected to investigate therapeutic targets against anoikis resistance.

4. Conclusions

In summary, we developed a versatile biomimetic composite hydrogel, GHP_4a, to construct a 3D cell spheroid in vitro, effectively mimicking the TME to investigate anoikis resistance. The GHP_4a2-5 hydrogel was synthesized by copolymerizing 4-arm-PEGDA, GelMA, and HAMA, achieving better biocompatibility, mechanical strength, and biomimetic activities. Crosslinking of 4-arm-PEGDA chains toughens the hydrogels, resolving the brittleness inherent in GelMA. Furthermore, GelMA contains RGD peptides that promote cellular interactions, while HAMA provides a hydrated, biomimetic environment that supports cellular functions. The above advantages together help overcome the limitations of using a single material. Our results showed that the multi-modal synergistic properties of the GHP_4a2-5 hydrogel recreate a physiologically relevant ECM-like environment, including a similar steady-state modulus of the human gut and the appropriate swelling, permeation, and degradation rates. This model effectively supports cell survival by mimicking natural adhesion and signaling mechanisms, specifically targeting key pathways such as FAK, PI3K/Akt, and caspase-mediated apoptosis, as indicated by the expression of specific markers (Bad, Ki-67, Mcl-1, PARP-1, and Survivin). Using the GHP_4a hydrogel as an ECM in a 3D culture of Caco-2 cells, we successfully developed an in vitro model to mimic the process of anoikis resistance in colorectal cancer, as demonstrated by immunohistochemical and transcriptomic analyses. It provides a robust tool for investigating therapeutic strategies targeting anoikis-resistant cell subpopulations and lays an essential biomimetic foundation for further exploring the metastatic mechanisms of colorectal cancer.

CRediT authorship contribution statement

Jia Weng: Writing – original draft, Investigation, Formal analysis, Data curation, Conceptualization. Shicheng Li: Writing – original draft, Methodology, Investigation. Jiacheng Weng: Methodology, Investigation, Data curation. Yan Wang: Validation, Supervision, Methodology, Investigation. Bincan Deng: Validation, Supervision, Methodology, Investigation. Mengxian Yao: Validation, Supervision, Methodology, Investigation. Hongke Hao: Validation, Supervision, Methodology, Investigation. Xia Huang: Validation, Supervision, Methodology, Investigation. Lei Gan: Writing – review & editing, Validation, Methodology, Investigation. Bo Chen: Writing – review & editing, Validation, Investigation, Funding acquisition, Formal analysis, Conceptualization. Xuan Xue: Writing – review & editing, Project administration, Funding acquisition, Conceptualization. Zhigang Chen: Writing – review & editing, Validation, Methodology, Funding acquisition, Formal analysis, Conceptualization.

Ethics approval and consent to participate

Animal management procedures were approved by the Animal Ethics Committee of Soochow University (number: 202312A0412) under the guidelines.

Declaration of competing interest

The authors declare that they have no conflicts of interest.

Appendix. ASupplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.