Extracellular matrix stiffness causes systematic variations in proliferation and chemosensitivity in myeloid leukemias

Abstract

Extracellular matrix stiffness influences biological functions of some tumors. However, it remains unclear how cancer subtypes with different oncogenic mutations respond to matrix stiffness. In addition, the relevance of matrix stiffness to in vivo tumor growth kinetics and drug efficacy remains elusive. Here, we designed 3D hydrogels with physical parameters relevant to hematopoietic tissues and adapted them to a quantitative high-throughput screening format to facilitate mechanistic investigations into the role of matrix stiffness on myeloid leukemias. Matrix stiffness regulates proliferation of some acute myeloid leukemia types, including MLL-AF9⁺ MOLM-14 cells, in a biphasic manner by autocrine regulation, whereas it decreases that of chronic myeloid leukemia BCR-ABL⁺ K-562 cells. Although Arg-Gly-Asp (RGD) integrin ligand and matrix softening confer resistance to a number of drugs, cells become sensitive to drugs against protein kinase B (PKB or AKT) and rapidly accelerated fibrosarcoma (RAF) proteins regardless of matrix stiffness when MLL-AF9 and BCR-ABL are overexpressed in K-562 and MOLM-14 cells, respectively. By adapting the same hydrogels to a xenograft model of extramedullary leukemias, we confirm the pathological relevance of matrix stiffness in growth kinetics and drug sensitivity against standard chemotherapy in vivo. The results thus demonstrate the importance of incorporating 3D mechanical cues into screening for anticancer drugs.

Keywords: biomaterials; cancer; drug screening; matrix stiffness; systems pharmacology.

Fig. 1.

Development of an integrative approach to systematically investigate the role of matrix mechanics in myeloid leukemias. (A) Schematic showing recapitulation of mechanical properties relevant to the hematopoietic system by ionic cross-linking of alginate hydrogels, followed by adaptation of the 3D hydrogels into quantitative screening and animal validation. (B) Different myeloid leukemia subtypes show distinct proliferative responses against matrix mechanics and ligand density. Ligand density is controlled by “degree of substitution” (DS), which indicates the number of RGD peptides conjugated per alginate molecule (0∼20). The whole cell population was used for viability analysis. The data were fit to biphasic dose–response curves for AML cells and standard dose–response inhibition curves for CML cells. *P < 0.05 from one-way ANOVA with Tukey’s honestly significant difference (HSD) test.

Fig. S1.

(A) Mechanical characterization of alginate hydrogels formed by different calcium concentrations (millimolar) during gelation. G′ (pascals) is the shear storage modulus. (B) IC₅₀ values for inhibition of cell proliferation across matrix stiffness by MK-2206. The IC₅₀ values were normalized against corresponding values on plastic. Error bars indicate ± SEM, n = 3 experiments, *P < 0.05 from one-way ANOVA with Tukey’s HSD test (relative to 0 Pa). (C) Levels of pAKT in myeloid leukemia cells across matrix stiffness were evaluated by intracellular flow cytometry. *P < 0.01 from one-way ANOVA with Tukey’s HSD test (relative to 0 Pa). (D) Dose–response curves of pAKT inhibition by MK-2206 in leukemia cells encapsulated in 300-Pa gels. pAKT was measured by intracellular flow cytometry. Values normalized to pAKT of cells in untreated 0 Pa.

Fig. 2.

Matrix stiffness regulates AML cell proliferation through autocrine signaling. (A) Simulations by a set of ordinary differentiation equations (SI Methods) mimic the biphasic cell proliferation of AML cells. (i) A modeling scheme showing an autocrine feedback circuit. α(c), rate of cell death as a function of soluble factor concentration; β(E), rate of cell proliferation as a function of E; δ(E), rate of soluble factor secretion as a function of E; γ, natural decay rate of soluble factors; c, soluble factor concentration; Leu, leukemia cells. The simulation results from (ii) increasing δE₅₀ alone and (iii) increasing both δE₅₀ and βE₅₀. The data were fit to biphasic dose–response curves. (B) MOLM-14 cells secrete factors that inhibit cell proliferation when cultured in 3D stiff gels. (Top) An experimental scheme. (Bottom) Total viable cell number after 7 d in the conditioned media from cells in different matrix stiffness. No Gel, 2D culture on plastic. n = 3 experiments, *P < 0.05 from one-way ANOVA with Tukey’s HSD test, 25 vs. 1,000 Pa. (C) Total leukemia cell numbers in 3D hydrogels with or without the presence of the AKT inhibitor MK-2206. The cell numbers from different conditions were normalized against that in the viscous matrix without drug treatment. The whole cell population was used for viability analysis. The data were collected from cells in alginate hydrogels with DS20 RGD. P on the x axis, 2D culture on plastic. n = 4 experiments, *P < 0.05 from one-way ANOVA with Tukey’s HSD test, control vs. MK-2206.

Fig. S2.

Developing a screen to quantify proliferation and pharmacodynamics. (A) An overview of the screen approach. Cells were transduced with mCherry-luciferase by lentivirus followed by single-cell mCherry positive clone selection, expansion, encapsulation of cells in gels, and deposition in a 96-well. Cell proliferation kinetics in the presence of different drugs was then evaluated by fluorescence. (B) Cell proliferation kinetics of select mCherry-positive K-562 clones. (C) The graph showing mCherry fluorescence vs. cell number of clone 3 cells encapsulated in 300-Pa gels. Linear fit R² > 0.95.

Fig. 3.

Matrix stiffness regulates drug sensitivity against distinct targets in myeloid leukemia subtypes. IC₅₀ values from (A) K-562 cells (clone 3, Fig. S2B) and (B) MOLM-14 cells (the whole cell population) treated with select drugs (for full name and target pathway of each drug, see Table S1) in 3D hydrogels conjugated with the RGD peptide (DS = 5) were normalized by respective IC₅₀ values from plastic, and then log-transformed before hierarchical clustering analysis. Drugs are classified into three classes: class I (ligand sensitive), class II (ligand and matrix stiffness sensitive), and class III (mechanics independent). The data were derived from n ≥ 15 experiments for A and n ≥ 4 experiments for B. Bold, underlined drugs belong to different classes in K-562 and MOLM-14 cells.

Fig. S3.

(A) IC₅₀ values of drugs from each drug class. For class II, one-way ANOVA P < 0.005, Tukey’s HSD test P < 0.05, correlation analysis P < 0.0001. (B) IC₅₀ values of drugs from all of the classes against K-562 cells across matrix mechanics. One-way ANOVA P < 0.05, followed by Tukey’s HSD test. (C) AUC values of dose–response curves across matrix mechanics for K-562 cells. One-way ANOVA P < 0.0001, Tukey’s HSD test ***P < 0.005, ****P < 0.001. In all graphs, values normalized to tissue culture plastic. (D) Clustering analysis of Hill slope values from individual drugs tested against K-562 cells.

Fig. S4.

(A) Drug resistance to class I and II drugs for K-562 (i) and MOLM-14 (ii) is integrin ligand-dependent. IC₅₀ values derived from cells in hydrogels with (DS = 5) or without RGD were normalized against plastic and plotted across different mechanics. Error bars indicate ± SEM, n = 3 experiments, *P < 0.05, paired t test. (B) Computational construction of a physical interaction network (SI Methods) reveals that proteins from protooncogenes of leukemias interact with distinct signaling clusters.

Fig. 4.

Leukemic oncogenes decouple the dependence of chemosensitivity on integrin ligands and matrix stiffness. (A) Overexpression of MLL-AF9 into K-562 cells increases their sensitivity against MK-2206 (AKT inhibitor) (i) but does not affect the sensitivity against sorafenib (RAF inhibitor) (ii) across different matrix stiffnesses. (B) Overexpression of BCR-ABL into MOLM-14 cells does not alter their sensitivity against MK-2206 (i) but increases the sensitivity against sorafenib (ii). The whole cell population was used for viability analysis. *P < 0.05, paired t test between empty oncogene vectors. n = 3 experiments. Error bars indicate ± SEM.

Fig. S5.

(A) Transduction of MOLM-14 cells with a retroviral vector containing empty (Left) or Bcr-Abl (Right) confirmed by flow cytometry analysis. A truncated form of NGFR was used as a reporter to both sort transduced cells using FACS and confirm transduction efficiency using flow cytometry. A red dotted line in the flow cytometry plots indicates cells that are not transduced with a NGFR vector. (B) Protein expression of Bcr-Abl and MLL-AF9 in cells by Western blot.

Fig. 5.

Resistance of leukemia cells against conventional chemotherapy in soft matrix in vivo. (A) Matrix stiffness affects K-562 (clone 3) cell growth in vivo. (i) Experimental scheme and representative images showing tumor growth from soft and stiff matrix in the human xenograft extramedullary leukemia model. (ii) Tumor growth for the first 3 wk after implantation described by first-order kinetics based on luminescence signals (normalized to Y0; see SI Methods). (iii) Mean initial growth rate. (iv) Mean deceleration rate. n = 15 mice, three experiments, *P < 0.01, paired t test. (B) CML cells are resistant to Ara-C in soft matrix. (i) Experimental scheme. (ii) Ara-C increases deceleration rate in stiff but not in soft matrix. (iii) Ara-C decreases tumor volume at week 6 in stiff but not in soft matrix. n = 6 mice, two experiments, *P < 0.01 from one-way ANOVA with Tukey’s HSD test, stiff control vs. treated.

Fig. S6.

(A) The number of viable cells after cell encapsulation in gels at day 0. (B) Slow but sustained growth of K-562 cells upon matrix stiffening. Matrix stiffness influences parameters of proliferation kinetics of K-562 cells. (i) Representative Gompertz curve fits (SI Methods) over 2 wk. (ii) Initial growth rate. (iii) Deceleration rate. For ii and iii, one-way ANOVA P < 0.05 with Tukey’s HSD test, *P < 0.05, **P < 0.01, ***P < 0.005. (C) K-562 cells in soft matrix show enhanced tumor growth compared with stiff matrix after s.c. implantation in vivo. Average radiance data from soft and stiff were converted to the natural log scale and divided (SI Methods). Straight line fit between day 3–14: Y = 0.084t + 0.70 (t: day). Data from n = 15 mice from three independent experiments. Paired t test, *P < 0.005. Error bars indicate ± SEM. (D) Tumor growth kinetics with Ara-C. (Plateau, acceleration rate) from the first-order kinetics fit for (i) stiff untreated (40, 0.56), Ara-C (20.8, 0.84) and (ii) soft untreated (22.5, 1.17), Ara-C (17.7, 1.15). Data from n = 8 mice from two independent experiments. Paired t test *P < 0.05. Error bars indicate ± SEM. (E) Bioluminescence signals become saturated at week 4. The same scale as in Fig. 4A was used. (F) Tumor volume at week 6 after implantation with different doses of Ara-C. Data from n = 5 mice for each group from two independent experiments. One-way ANOVA with Tukey’s HSD test *P < 0.05 control vs. 62.5mg/kg Ara-C. Error bars indicate ± SEM.

Fig. S7.

Histological sections from s.c. implanted K-562 cells stained with hematoxylin and eosin. (A) A representative image showing blood cells (blue), stromal-like cells (green), and hydrogel fragments (yellow) after 2 wk of tumor implantation. (B) Representative images showing general morphological features of different treatment groups. Yellow: A region with nuclear fragments. (Scale bars: 50 μm.)

Fig. S8.

Immunofluorescence staining of histological sections from s.c. implanted K-562 cells. (A) Representative images showing nuclei (blue) and cleaved caspase-3 (red). (Scale bar: 50 μm.) (B) Representative images showing F-actin (red), human mitochondria (green), and YAP (cyan). (Scale bars: 50 μm.)