A structurally tunable DNA-based extracellular matrix

The construction of artificial matrices that mimic the structure and function of the naturally occurring extracellular matrix (ECM) is a central challenge for tissue engineering.¹ The difficulty lies in our inability to replicate the biophysical and chemical signaling effectors found in the ECM. Although polymeric hydrogels, nanofibrillar synthetic biomaterials, and PEG-based materials have been used for cell adhesion and proliferation,² there is a need for scaffolds that are structurally programmable. DNA nanotechnology uses base-pairing to generate predefined one-, two- and three-dimensional assemblies for the templated growth of materials, the organization of proteins and nanoparticles, and as platforms for diagnosis and delivery.³ Here we combined the principles of DNA nanotechnology with protein engineering to generate a new class of DNA/protein extracellular matrices, and demonstrated their potential for cell adhesion and growth. The use of DNA means that our assemblies are inherently amenable to structural programming, which we demonstrated and used to fine-tune cell morphology, behavior, signaling, and transcription factor localization.

A DNA/protein-based matrix (ECM_DP) was constructed from a DNA ribbon (ECM_D) surface-functionalized with proteins (ECM_P) containing the RGD-domain of human fibronectin. This protein was selected because of its role in cellular adhesion, survival, proliferation and migration.⁴ The structure of ECM_D was inspired by collagen. It is assembled from five DNA strands that form a 4-helix ribbon^3d with overhangs (Scheme 1a), and analyzed using atomic force microscopy (AFM; Figure 1a).¹¹ ECM_P was expressed in P. pastoris, and contains a FN_III10 domain, a flexible (Gly₄Ser)₃ linker, a monomeric streptavidin domain for conjugation to 5′-biotin DNA, and a His₆ tag (Scheme 1b).¹¹ The 1:1 binding of ECM_P with DNA was confirmed using gel-shift assays.¹¹ Addition of ECM_P to ECM_D generated the DNA/protein matrix ECM_DP (Figure 2b).¹¹

Scheme 1.

Figure 1.

1–6 correspond to plates coated with ECM_DP, ECM_D, ECM_P, unconjugated ECM_D and ECM_P, polylysine, and fibronectin, respectively. (a) AFM characterization of ECM_D, with cross-sectional height analysis (lower panel). Bar 10 μm. (b) AFM characterization of ECM_DP. Bar 1 μm. (c) HeLa cells on ECM_DP-coated glass plates. Bar 15 μm. (d) Percentage of cells that attach onto ECM_DP-coated plates, and onto 2–6. (e) Rate of attachment. (f) Strength of attachment, as a function of detachment. (g) Viability of cells growing on ECM_DP. (h) Migration of HeLa cells growing on ECM_DP-coated plates.

Figure 2.

(a) Single-stranded domains are used to fine-tune scaffold persistence length. (b) AFM analysis ECM_DP, ECM_DP3, ECM_DP10. (c) Cell morphology as a function of single-stranded character. Red, actin. Blue, nucleus. (d) Relative amounts of integrin receptors bound to ECM_DP, ECM_DP3 and ECM_DP10 (lanes 1–3, respectively). (e) Relative p-FAK expression levels as a function of ECM_DP, ECM_DP3 and ECM_DP10 (lanes 1–3, respectively). (f) GFP-FOXO1a localization as a function of scaffold persistence length. Statistics (right).

ECM_DP promoted cell adhesion and migration. HeLa cervical cancer cells were used because of their well-characterized behavior,⁵ and were found to attach onto ECM_DP-coated glass plates with 96% efficiency (Scheme 1c, Figure 1c, 1d).¹¹ Control experiments included using plates coated with ECM_D, ECM_P, unconjugated ECM_D and ECM_P, polylysine, or fibronectin.⁶ These controls reveal that HeLa cells were incapable of effectively adhering onto ECM_D or ECM_P, unless assembled into ECM_DP, and that ECM_DP was as effective as polylysine or fibronectin for cell adhesion (Figure 1d).¹¹ The half-time of cell adhesion onto ECM_DP was 6 minutes, which was considerably faster than their rate of attachment onto polylysine or fibronectin (Figure 1e).¹¹ The force of adhesion was determined as a measure of detachment. We pretreated cells with the protein synthesis inhibitor cycloheximide to prevent deposition of cell-derived ECM proteins. As seen in Figure 1f, 90% of the cells remain attached to ECM_DP when a detachment force of 780 pN was applied,¹¹ which was considerably greater than the percentage of cells that remained attached to either polylysine or fibronectin. We also compared the effectiveness of ECM_DP to matrigel-coated surfaces, and found it to be comparable.¹¹ ECM_DP can thus be used as an effective extracellular matrix for cell seeding and attachment.

HeLa cells attached to ECMD_P remained viable and migrated. We conducted flow cytometry experiments on cells stained with the apoptosis and necrosis fluorophores Annexin-V-EGFP and 7-AAD⁷. Only 4% of cells attached to ECM_DP stained positive for the apoptosis marker 48 hours after adhesion, and none were necrotic (Figure 1g).¹¹ In contrast, most of the cells attached to ECM_D died,¹¹ indicating that the protein element of ECM_DP was essential for cell survival. Motility was determined using a modified Oris^™ cell-migration assay. HeLa cells moved on ECM_DP at a rate of 46μm/day (Figure 1h), which was slightly faster than their rates onto polylysine- or fibronectin-coated plates.¹¹ ECM_DP thus promotes cell survival and migration.

The structure of the ECM influence the cell’s cytoskeletal arrangement and shape.⁷ ECM_DP is fully double-stranded with a persistence length of 1.2 μm (i.e. length at which bending starts).¹¹ The incorporation of single-stranded domains should result in the formation of increasingly flexible assemblies that are structurally different (Figure 2a). ECM_DP3 and ECM_DP10 possess 3 and 10 base single-stranded regions and have persistence lengths of 0.95 and 0.32 μm, respectively (Figure 2b).¹¹ ECM_DP3 and ECM_DP10 are structurally more flexible than the fully double-stranded ECM_DP. Importantly, HeLa cells growing on EMC_DP, ECM_DP3 and ECM_DP10 were progressively more rounded (Figure 2c). These differences were confirmed by quantification of the substrate-bound integrin receptors (Figure 2d).¹¹ Focal adhesions form in response to increasing ECM stiffness.⁷ Vinculin immunohistochemistry revealed a relative decrease in the number of assembled focal adhesions with increasing ECM flexibility.¹¹ We confirmed that the observed differences in cellular behavior is due to the structural changes resulting from the use of single-stranded DNA domains, and confirmed this engineering a set of fully double-stranded DNA assemblies analogues to ECD_DP, but with protein-to-protein distances that are identical to ECM_DP3 and ECM_DP10.¹¹ Collectively these results indicate that a profound cytoskeletal response can be achieved by controlling substrate structure. Using this, we proceeded to manipulate a number of integrin-mediated processes, such as the status of transduction proteins and the localization of intracellular transcription factors.

Focal adhesion kinase (FAK) is a signal transduction protein involved in cell development, migration, and in the regulation of the tumor suppressor p53.¹⁰ FAK is phosphorylated in response to focal adhesion formation,¹⁰ and can thus be fine-tuned as a function of ECM stiffness. As seen in Figure 2e, we were able to modulate the intracellular levels of p-FAK by modulating the degree of ECM single-stranded character.¹¹

The integrin pathway can be used to activate the cytoplasmic versus nuclear localization of the transcription factor FOXO1a.¹⁰ To test whether ECM stiffness could be used to fine-tune the intracellular localization levels of FOXO1a, we engineered a stable human osteosarcoma U2OS cell line expressing GFP-tagged FOXO1a fusions.¹¹ As seen in Figure 2f, by increasing the ECM’s single-stranded character, we were able increase nuclear levels of FOXO1a.¹¹ We modulated FOXO1a localization, with potential consequences for cell differentiation, proliferation, longevity, and apoptosis.⁹

In summary, we constructed a new class of artificial extracellular matrices. Our approach involved the assembly of one-dimensional DNA ribbons decorated with cytoskeletal protein elements. We demonstrated the potential of this material for ex vivo cellular scaffolding by showing that cells attach, survive and grow, and demonstrated its capacity for structural programmability by fine-tuning cell morphology, cytoskeletal organization, signal transduction, and transcription factor localization. The diversity and addressability of DNA nanostructures suggests that a multitude of artificial extracellular matrices could in principle be assembled using this approach. This contribution thus lays the groundwork for the modular construction of programmable ECMs to systematically study and replicate the naturally occurring extracellular niche.