Formation of Multi-Component Extracellular Matrix Protein Fibers

The extracellular matrix (ECM) consists of polymerized protein monomers that form a unique fibrous network providing stability and structural support to surrounding cells. We harnessed the fibrillogenesis mechanisms of naturally occurring ECM proteins to produce artificial fibers with a heterogeneous protein makeup. Using ECM proteins as fibril building blocks, we created uniquely structured multi-component ECM fibers. Sequential incubation of fibronectin (FN) and laminin (LAM) resulted in self-assembly into locally stacked fibers. In contrast, simultaneous incubation of FN with LAM or collagen (COL) produced molecularly stacked multi-component fibers because both proteins share a similar assembly mechanism or possess binding domains specific to each other. Sequential incubation of COL on FN fibers resulted in fibers with sandwiched layers because COL molecules bind to the external surface of FN fibers. By choosing proteins for incubation according to the interplay of their fibrillogenesis mechanisms and their binding domains (exposed when they unfold), we were able to create ECM protein fibers that have never before been observed.

with PSS micro-patterns ( Fig. 1g; see Methods section). Negatively charged synthetic polymer or phospholipid surfaces have been previously used to replicate the negative surface charge of PGs, enabling the extension and polymerization of FN or LAM molecules into large fibrillar networks [22][23][24][25][26][27][28][29][30] . Recently, we observed large-scale self-organizing single ECM protein networks on PG-mimicking PSS 28 . PSS works as a mimetic of PG because it possesses a functional group in common with PG: a sulfonyl group. 29,30 The sequence of negatively charged groups in PSS attracts the positively charged domains in ECM proteins and induces protein unfolding, leading to fibrillogenesis. Only β-sheet-rich ECM proteins (FN and LAM) formed a long-ranged fibrillar structure; β-sheetpoor ECM proteins (COL and ELAS) did not.

Results
We discovered that the negatively charged PSS surface induced protein unfolding of FN and LAM and subsequent self-assembly into fiber networks but did not have the same effect on COL fibers or adsorbed ELAS fibers, which instead were randomly distributed (see Supplementary Fig. S1a-f). These inferences were verified by comparing the amide I peaks (which are highly sensitive to changes in protein secondary structures 31 ) for the IR images of native FN and LAM and fibers induced on a spin-coated PSS surface ( Fig. 1a-f). The amide I peaks of FN and LAM fibers were red-shifted (Fig. 1c,f), indicating that both proteins unfolded upon binding to a charged domain 28,32,33 , but the amide I peaks of the randomly distributed COL fibers and ELAS adsorbates revealed no red shift (see Supplementary Fig. S1c,f). These different fibrillogenesis characteristics enable diverse in vivo combinations of composite ECM protein networks. FN and LAM were also found to form controlled networks among PSS islands (see Supplementary Fig. S1,g,h), while COL and ELAS formed random fibers unrelated to the islands (see Supplementary Fig. S1,i,j). These structures reflect the natural tendencies of different ECM proteins to form fibers based on the presence or absence of a surface charge.
When LAM or FN were sequentially incubated with the FN fibers (red, Fig. 2a-c and see Supplementary Video 1) or LAM fibers (red, Fig. 2d,f and see Supplementary Video 2), respectively, they were locally stacked on the pre-existing fibers. The magnified z-projected images (Fig. 2c,f) show that pre-incubation, proteins underwent self-induced fibrillogenesis to form fibrils, and that post-incubation, the new fibrils moved into empty spaces between the pre-existing protein's fibrils, where they locally self-assembled to make continuous protein composite fibers, which we termed "locally stacked composite fibers. " Surprisingly, simultaneous incubation of a mixture of FN and LAM formed composite FN/LAM fibers ( Fig. 2g-i and see Supplementary Video 3). The homogeneous distribution of both proteins in these composite fibers (Fig. 2i) is not seen in locally stacked fibers and clearly indicates that FN and LAM fibrils were hetero-stacked during the initial spontaneous fibrillogenesis. We termed these hybridized fibers "hetero-stacked composite ECM fibers. " The hetero-stacking process results from chemical binding during the initial spontaneous fibrillogenesis, in contrast to the process that creates locally stacked fibers, which involves the insertion of non-hybridized self-assembled protein fibrils into pre-existing fibers.
The results demonstrated in Fig. 2 led us to inquire whether we could produce other types of composite ECM fibers by harnessing the protein-binding domains in ECM proteins, even when the proteins' fibrillogenesis mechanisms differ. To address this question, we incubated COL and ELAS with pre-existing FN fibers because FN contains a COL-binding domain (I 6-9 ) 34 but not an ELAS-binding domain. When COL was sequentially incubated with pre-existing FN fibers, the FN fibers were coated with COL ( Fig. 3a-c and see Supplementary Video 4) and formed a third type of composite ECM fiber, which we termed a "sandwiched composite fiber. " COL molecules accumulated on the surface of the FN fibers because they are sticky and strongly interact with suspended FN fibers.
In contrast, simultaneous incubation of FN and COL on the PSS patterns formed hetero-stacked composite fibers (Fig. 3d-f and see Supplementary Video 5) because COL cannot by itself form ordered networks on PSS micro-patterns (see Supplementary Fig. S1,i). Similar to the findings shown in Fig. 2g-i, both proteins were irregularly distributed throughout the fiber (Fig. 3i). As the FN fibers were induced on the PSS micro-patterns, FN molecules unfolded to expose COL binding sites, which led to the binding of COL to FN and initiated simultaneous self-induced assembly of the molecules into a single fiber.
As a negative control, ELAS was sequentially incubated on the FN substrate. The ELAS fibers were randomly adsorbed on the surface without interacting with the FN fibers ( Fig. 3g-i, see Supplementary Fig. S1,j, and Videos) because ELAS fibrillogenesis relies neither on charge-initiated unfolding nor on FN-induced assembly. The fluorescence intensity profile in Fig. 3i does not reveal any interaction between FN and ELAS.

Discussion
Using both sequential and simultaneous incubations (Figs 2 and 3), we successfully engineered three types of composite ECM fibers: locally stacked composite fibers, hetero-stacked composite fibers, and sandwiched composite fibers (Fig. 4). Three-dimensionally (3D) reconstructed confocal images (Fig. 4a,b) show that sequential incubation of FN and LAM with pre-existing fibers formed locally patched fibers. Both the 3D-reconstructed z-stack images and the SEM images (Fig. 4h) of these structures show that binding domain-induced fibrils continuously assembled on the pre-existing fibers. Locally patched stacked were thicker than single-component ECM fibers due to the accumulation of protein fibrils (Fig. 4g).
In contrast, simultaneous incubation of FN and LAM on PSS micro-patterns produced hetero-stacked (i.e., β-sheet-β-sheet) composite ECM fibers because the β-sheet-rich FN and LAM underwent synchronized fibrillogenesis on the micro-patterns. The mechanism of the LAM/FN hetero-stacking during their simultaneous adsorption in a single fiber is obscure. It has been known, however, that the pre-existing LAM helps the self-assembly of FN 12 , then the charged-induced LAM or FN might mutually interact with FN or LAM molecules via the association of β-sheets in formation of the fibrils. Likewise, simultaneous incubation of FN and COL on PSS micro-patterns produced hetero-stacked (i.e., α-helix-β-sheet) composite ECM fibers. Because COL possese neither β-sheet in its secondary structure nor an unfolding mechanism, it cannot perform charge-initiated unfolding and self-assembly on PSS micro-patterns (see Supplementary Fig. S1,i). However, COL and FN can bind to each other and undergo synchronized fibrillogenesis because FN has a binding domain for COL 7 . While β-sheet rich FN and LAM molecules can mutually interact with each other, the helical COL molecules bound only to the exposed FN binding sites. Nevertheless, multiple FN domains in a single FN molecule allow the formation of insoluble fibers even with the structural interference of bound LAM or COL. The thickness of hetero-stacked fibers is similar to the thickness of locally stacked fibers owing to the homogeneous distribution of the constituent proteins in both cases (Fig. 4g).
Lastly, when COL was incubated with pre-existing FN fibers, COL covered the surface of the FN fibers owing to the specific FN-COL binding interaction (Fig. 4e,f). An SEM image (Fig. 4i) shows that COL fibers were adsorbed, not stacked, on the surface of the FN fibers. These fibers were significantly thicker than both single-component fibers and composite fibers because COL molecules accumulated on the outside surface of the FN fibers (Fig. 4g).
Deciphering the influence of different cell binding domains of the multi-composite ECM protein fibers on cell response is of great importance, in terms of biological and clicnical implications for the tissue specific ECM environment. We have confirmed that the PSS-induced FN fibers possessed excellent cell viability and resulted in functional tissue formation 28 . Subsequent and separate studies will further reveal how the binding sites and compositions of the multiple component fibers affect cell fate.
To the best of our knowledge, this is the first report of the production of multi-component ECM fibers via simultaneous or sequential incubations of ECM proteins chosen for their specific fibrillogenesis mechanisms and binding domains. Sequentially or simultaneouly incubating together any two of the FN, LAM, or COL proteins formed one of three different types of multi-component protein fibers: (1) locally stacked composite ECM fibers, (2) hetero-stacked composite ECM fibers, or (3) sandwiched composite ECM fibers. We found that ECM proteins could be used as fundamental units-like polymerizable monomers-to synthesize new multi-component fibers. We project that researchers will be able to use the fibrillogenesis mechanisms and binding domains in a variety of ECM proteins from diverse parts of the human body to create heterogeneous ECM architectures.

Preparation of ECM fibrillogenesis on spin-coated PSS surfaces and FT-IR analysis.
Poly(styreneco-4-styrene sulfonic acid) (33% PSS [mol% of SO 3 H: 33], Polymer Source, USA) was dissolved in dimethylformamide (Sigma Aldrich, Korea). For FTIR imaging, 33% PSS solutions were spin-coated onto gold-coated silicon wafers (Sigma) at 2000 rpm for 1 min. The PSS-coated surfaces were annealed at 150 °C overnight. After annealing, the surfaces were incubated with 66 µg/ml FN (Invitrogen, USA), LAM (Invitrogen, USA), type-Ι COL (Sigma Aldrich, Korea), or ELAS (Sigma Aldrich, Korea) in phosphate-buffered saline (pH 7.4) for 72 h at 37 °C. After incubation, the samples were washed with deionized water and dried in a desiccator. FT-IR images of the dried samples were acquired from the collected spectra, which were obtained using a Cary 660 spectrometer and a Cary 620 FT-IR microscope (Agilent Technologies, Korea). The resolution and total number of scans of the spectra were 4 cm −1 and 32, respectively, in the 800-4,000 cm −1 wavenumber regions with Agilent's 16 × 16 focal plane array. Note that the background was corrected for each measurement. Preparation of single-and multi-component ECM fibers on PSS micro-patterns, sequential or simultaneous protein incubation, and confocal microscopy. PSS micro-patterns were prepared using the micro-contact printing technique described in our previous study 28 . Briefly, square arrays with star-like patterns were prepared by photolithography and soft lithography, and 33% PSS solution was inked onto polydimethylsiloxane (PDMS) stamps and then imprinted on coverslips. ECM proteins and polyvinyl acetate were sequentially or simultaneously incubated on the 33% PSS-patterned coverslips for 6 h and 72 h, respectively. After the final incubation, the samples were mounted on a microscope stage, and z-stacked images were obtained using a Zeiss LSM 5 LIVE confocal microscope. Image analysis. The z-stacked images were analyzed with Image J software (NIH). Z-projected images were obtained by the Image J volume viewer 35 . Three-dimensional animations of z-stacked images were compiled by the Image J 3D viewer 36,37 . Data availability. All data generated or analyzed during this study are included in this article (and its supplementary information files).