A novel fibronectin-binding peptide reveals dynamic intermediate structures during matrix assembly Authors: Henry A. Resnikoff, Aaron J. Hamlin, Faith M. Reash, Emily N. Sabatino, Isabella M. Strumolo, Luqiong Wang, Jean E. Schwarzbauer Published: 2025 ABSTRACT The extracellular matrix (ECM) is an expansive network of polymers that regulates cell adhesion, migration, and tissue morphogenesis, and that becomes dysregulated in fibrotic disease states. Of the many ECM proteins, fibronectin (FN) is foundational and FN fibrils are instrumental in promoting deposition of many other ECM proteins. FN fibrils form by cell receptor-mediated polymerization of secreted FN dimers. To provide a deeper understanding of the FN assembly process, we sought to identify FN-binding peptides that could be used to target FN during matrix assembly. Here we report the isolation of a FN-binding peptide, S2, by M13 phage display. When displayed within a GST-S2-EGFP fusion protein, S2 directs incorporation of the fluorescent fusion protein into FN fibrils during fibroblast matrix assembly, enabling direct visualization of fibril formation and accumulation. The S2 fusion protein remains stably associated with FN in decellularized matrices. Live cell imaging with GST-S2-EGFP revealed events in matrix formation and maturation and highlighted distinct features of FN assembly such as nascent fibril nucleation and elongation from coalescing FN aggregates. This S2 peptide fusion protein represents a new tool for real-time analysis of the ECM, for generation of fluorescent scaffolds, and for directing functional proteins to FN matrix.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-33556-9. Introduction Fibronectin (FN) is a foundational extracellular matrix (ECM) molecule that plays essential roles in cell adhesion, migration, and differentiation1. It serves as a scaffold for cells and other ECM proteins, influencing tissue organization and cellular signaling2–4. FN is essential for embryonic development5 and plays key roles in tissue morphogenesis6–8. Dysregulation of FN assembly has been implicated in a wide range of pathological conditions including fibrosis9–12, cancer13,14, skeletal abnormalities15,16 and cardiovascular disease17. FN exists as a large glycoprotein composed of multiple functional domains that facilitate interactions with integrins, growth factors, and other ECM molecules18,19. It is secreted as a soluble dimer and undergoes fibrillogenesis, a process driven by cell-FN interactions that converts soluble FN into insoluble fibrils4,19. This dynamic assembly is crucial for ECM deposition and organization. Despite its importance, visualizing FN matrix formation in live-cell systems remains a technical challenge, limiting our ability to study its real-time dynamics and cellular interactions. Traditional methods for studying FN matrix include immunofluorescence staining and addition or expression of fluorescently-tagged proteins. While these approaches provide valuable insights, they each have drawbacks for studying ECM dynamics. For example, immunofluorescence staining often requires fixation, which halts the assembly process, whereas genetically-tagged ECM proteins limit study of FN to the organisms/cells in which the tag exists. Addition of fluorescently-tagged FN to cell cultures marks sites where assembly is occurring but may give incomplete labeling that may bias analysis. As such, a probe that could be deployed in diverse systems and allow for live imaging of FN would facilitate analysis of FN matrix assembly and help further our understanding of FN in both healthy and pathological contexts. Adherent cultured cells such as fibroblasts or endothelial cells deposit FN-rich matrices that mimic aspects of tissue microenvironments. These and other in vitro models have been instrumental in understanding FN’s role in physiological and pathological conditions. FN visualization through super-resolution microscopy20 and molecular probes that differentially stain stressed versus relaxed FN fibrils21,22 have advanced our understanding of specific features of the ECM. A tool that stably incorporates into FN fibrils would be very useful for studying FN matrix assembly, remodeling, and cell-matrix interactions in static and real-time situations. To address this goal, we leveraged phage display technology to develop a novel GFP-tagged FN-binding protein that integrates into FN fibrils. We identified a unique peptide that allows for live imaging of FN matrix formation and offers the unique advantage of matrix retention after decellularization, enabling the study of fluorescent FN matrices in both culture systems and cell-derived scaffolds. Here, we describe the development and application of this tool for visualizing FN matrix dynamics in both fixed and live-cell environments. Results Identification of FN-binding peptides using phage display We screened an M13 phage display library for peptides that bind to a recombinant N-terminal 70 kDa fragment of human FN. The 70 kDa fragment is composed of two domains, the assembly domain with the FN-binding site that is required for matrix assembly and the gelatin/collagen-binding domain (Supplementary Figure S1). The phage display library displaying 12 amino acid-long peptides was panned against plates coated with the 70 kDa fragment following the workflow in Supplementary Figure S1. After 3 rounds of panning, 47 phages were randomly selected for sequencing, which yielded 26 unique peptide sequences. 11 of the 26 sequences were found in multiple phage isolates (Fig. 1A). Binding of individual phage clones to 70 kDa was confirmed with an enzyme linked immunosorbent assay (ELISA), using pooled unbound phage from the first pan as a negative control. All but 2 of the 26 unique sequences generated a positive ELISA signal, as defined by a mean signal higher than 3x the average negative control signal (Fig. 1A). Fig. 1FN-binding peptides identified by phage display. (A) M13 peptide phages were screened against surfaces coated with the 70 kDa fragment of human FN. Phages arbitrarily numbered S1 through S47 were used for DNA sequencing. Each unique peptide sequence and the phage clone numbers are listed (left and middle columns). Note that some sequences were present in more than one phage isolate. Phage binding to 70 kDa was confirmed in an ELISA, with “+” indicating a signal at least 3 times the negative control. (B) NIH 3T3 cells grown on glass coverslips were fixed, blocked with BSA, and then incubated with 1 × 1011 pfu/mL phage for 1 h before immunostaining for FN (left, green) and M13 phage protein (middle, red) with the merged images shown on the right. Examples show representative photomicrographs of three phages and depict two phages (S2, S12) that demonstrated positive binding and S33 which did not bind. Scale bar = 200 μm. FN-binding peptides identified by phage display. (A) M13 peptide phages were screened against surfaces coated with the 70 kDa fragment of human FN. Phages arbitrarily numbered S1 through S47 were used for DNA sequencing. Each unique peptide sequence and the phage clone numbers are listed (left and middle columns). Note that some sequences were present in more than one phage isolate. Phage binding to 70 kDa was confirmed in an ELISA, with “+” indicating a signal at least 3 times the negative control. (B) NIH 3T3 cells grown on glass coverslips were fixed, blocked with BSA, and then incubated with 1 × 1011 pfu/mL phage for 1 h before immunostaining for FN (left, green) and M13 phage protein (middle, red) with the merged images shown on the right. Examples show representative photomicrographs of three phages and depict two phages (S2, S12) that demonstrated positive binding and S33 which did not bind. Scale bar = 200 μm. Evaluating peptide-FN interactions To determine whether the peptides can bind to FN fibrils, we used NIH 3T3 fibroblasts which assemble a dense fibrillar FN matrix. Cultures were fixed and incubated with individual phage clones, followed by co-staining with antibodies against FN and an M13 phage protein. Phage binding was scored as positive if phage signal both resembled and overlayed the fibrillar structure of the FN matrix (Fig. 1B, Supplementary Figure S2). Three of the phage clones (S2, S12, S39) exhibited a fibrillar pattern with extensive colocalization with the FN matrix. The other M13 phage staining resembled the negative control (wild-type M13 phage, WT), with very little phage signal shown in two examples (S33 and S37). We conclude that while most of the peptides can target phages to FN-coated plastic (Fig. 1A), only 3 peptides showed the capacity to bind phage to FN fibrils in this assay. The observation that most of the peptides did not bind to FN fibrils after assembly and fixation suggests that the binding sites might not be accessible or may be affected by fixation. Phage clones were incubated with growing fibroblast cultures during matrix assembly for 48 h followed by fixation and staining. Since our phage display screening was against human 70 kDa, we used human WI-38 fibroblasts in this experiment to assess incorporation of phage into FN matrix. As with the matrix binding experiment with mouse cells, phages with peptide S2, S12, or S39 showed significant incorporation into the human FN matrix (Fig. 2A). The WT negative control phage showed very little association with FN fibrils. In contrast to the binding results, a number of phage peptides showed signal above background in the incorporation experiment, such as S37 phage (Fig. 2). To quantify the relationship between FN and phage signals, we generated Pearson’s correlation coefficients, which measure how closely pixel intensities in one channel correlate with those in the other (range − 1 to + 1). Values near + 1 indicate strong positive colocalization, 0 indicates no linear relationship, and − 1 indicates inverse correlation. Only 4 of the phage signals had Pearson’s coefficients higher than 0.5. S2, S12, and S39 had the highest coefficients (0.70, 0.61, 0.57, respectively; Fig. 2B). S37 was also higher than the others (0.53). However, most of the phages had coefficients below 0.5 indicating limited fibril-phage overlap. Fig. 2Phage incorporation into FN matrix assembled by growing fibroblasts. Human WI-38 fibroblasts were seeded and allowed to attach before addition of individual phage clones to the media (1 × 1011 pfu/mL). Cells with phage were grown for 2 days and then immunostained for FN (green) and phage (red). (A) Representative images are shown depicting FN, phage, and merge channels for clones S2, S12, S33, S37, S39, and WT control. Scale bar = 200 μm. (B) Correlation between FN and phage staining was determined by calculating Pearson’s correlation coefficients from three views of each sample for the indicated phages. Dotted line marks coefficient of 0.5. Phage incorporation into FN matrix assembled by growing fibroblasts. Human WI-38 fibroblasts were seeded and allowed to attach before addition of individual phage clones to the media (1 × 1011 pfu/mL). Cells with phage were grown for 2 days and then immunostained for FN (green) and phage (red). (A) Representative images are shown depicting FN, phage, and merge channels for clones S2, S12, S33, S37, S39, and WT control. Scale bar = 200 μm. (B) Correlation between FN and phage staining was determined by calculating Pearson’s correlation coefficients from three views of each sample for the indicated phages. Dotted line marks coefficient of 0.5. Differences between phage in their colocalizations with FN signals could result from concentration differences between amplified phage stocks. We decided to create high concentration phage stocks for the two phages with the highest Pearson’s coefficients, S2 and S12, to see if there was a dose-dependent effect. Incubation with a 10-fold higher concentration of S2 phage gave essentially complete coverage of fibrils by phage (Fig. 3A-C, G) while S12 phage gave increased coverage but also showed an increase in punctate background signal (Fig. 3D-F). The Pearson’s coefficient for S2 (0.82) was also increased, while the Pearson’s coefficient for S12 did not indicate greater colocalization with more concentrated phage (0.55). Altogether, we conclude that the S2 phage gives highly co-localized FN fibril labeling compared to the other peptides. This conclusion is further supported by confocal images depicting a single fibril within an NIH 3T3-derived ECM in which there is significant overlap between phage and FN signals (Fig. 3G, H, I). Fig. 3S2 phage at high concentration demonstrates specific binding to FN fibrils. Clones S2 (A - C) and S12 (D – F) at 1 × 1012 pfu/mL were incubated with NIH 3T3 fibroblasts for 2 days. Assembled matrices were fixed and stained for FN (green) and phage (red). Representative images are shown. (G) A merged maximum intensity projection obtained by confocal microscopy from the S2 sample is shown (FN green, phage red). A single fibril from this field (outlined by a blue rectangle) was captured at higher magnification and shown in H (FN) and I (phage). Pearson’s coefficients were calculated from multiple fields, averaged, and depicted in the bottom right corner of C and F. Scale bar = 100 μm. S2 phage at high concentration demonstrates specific binding to FN fibrils. Clones S2 (A - C) and S12 (D – F) at 1 × 1012 pfu/mL were incubated with NIH 3T3 fibroblasts for 2 days. Assembled matrices were fixed and stained for FN (green) and phage (red). Representative images are shown. (G) A merged maximum intensity projection obtained by confocal microscopy from the S2 sample is shown (FN green, phage red). A single fibril from this field (outlined by a blue rectangle) was captured at higher magnification and shown in H (FN) and I (phage). Pearson’s coefficients were calculated from multiple fields, averaged, and depicted in the bottom right corner of C and F. Scale bar = 100 μm. Generation and testing of a peptide-EGFP fusion protein Phage clones were useful for characterizing peptide-FN binding activities and confirming FN fibril localization. However, there is an upper limit to the phage concentrations that can be easily generated and for some experiments, increasing the phage concentration led to higher background staining. Therefore, we decided to create a peptide fusion protein. We selected the pGEX4T3-GST-EGFP expression plasmid, into which the peptide cDNA was inserted between glutathione-S-transferase (GST) and green fluorescent protein (EGFP) sequences yielding the GST-S2-EGFP fusion protein (Fig. 4A). A tobacco etch virus (TEV) protease cleavage site lies between GST and the peptide, enabling removal of the GST tag. Thus, the peptide fusion protein can be purified via the GST tag and subsequently cleaved to produce a peptide-EGFP protein. Fig. 4GST-S2-EGFP fusion protein retains FN fibril binding capacity. Left: Cartoons depict the compositions of GST-peptide-EGFP fusion proteins created with S2 and S3 peptides (A, D), the control protein derived from the parent vector with the peptide SNSRVD (B), and S2-EGFP after TEV cleavage of GST-S2-EGFP (C). Right: NIH 3T3 cell monolayers were incubated with individual fusion proteins added to the culture media at 5 µg/ml during 2 days of growth and then fixed and stained for FN (red). GFP signal in the matrix was detected by innate fluorescence. Representative images are shown from each of the fusion protein conditions. Only the intact GST-S2-EGFP protein bound to FN fibrils. Scale bar = 200 μm. GST-S2-EGFP fusion protein retains FN fibril binding capacity. Left: Cartoons depict the compositions of GST-peptide-EGFP fusion proteins created with S2 and S3 peptides (A, D), the control protein derived from the parent vector with the peptide SNSRVD (B), and S2-EGFP after TEV cleavage of GST-S2-EGFP (C). Right: NIH 3T3 cell monolayers were incubated with individual fusion proteins added to the culture media at 5 µg/ml during 2 days of growth and then fixed and stained for FN (red). GFP signal in the matrix was detected by innate fluorescence. Representative images are shown from each of the fusion protein conditions. Only the intact GST-S2-EGFP protein bound to FN fibrils. Scale bar = 200 μm. We tested the S2 fusion protein by adding it to fibroblast cultures to determine if it behaved similarly to the peptide-phage. We found that full-length GST-S2-EGFP fusion protein was incorporated into FN fibrils during assembly and showed significant overlap with FN immunostaining following fixation (Fig. 4A). A GST-S2-mScarlet fusion protein gave essentially the same FN binding and matrix incorporation results as GST-S2-EGFP. The specificity of GST-S2-EGFP for FN was shown with a solid phase binding assay. The fusion protein bound to the N-terminal 70 kDa fragment and to full length FN, but did not bind to other FN fragments or to other ECM proteins (laminin-111, fibrinogen, and type I collagen) (Supplementary Figure S3). The GST-Control-EGFP fusion protein did not associate with FN fibrils (Fig. 4B) indicating that the association is not due to the fusion partners. The GST tag was cleaved to yield S2-EGFP, but cultures incubated with this truncated fusion protein showed no fluorescent signal (Fig. 4C). We also found no S2-EGFP incorporation after incubation with an equimolar mixture of GST and S2-EGFP proteins (not shown). We conclude that the cleaved fusion protein is unable to be incorporated into FN fibrils. A fusion protein was generated using peptide S3, which supported phage binding to 70 kDa in a solid phase assay (Fig. 1A) but only weak association with FN matrix (Fig. 2B). GST-S3-EGFP showed no detectable FN fibril binding in culture (Fig. 4D). Together, these data demonstrate that the S2 peptide is necessary for binding of GST-S2-EGFP to FN fibrils and that binding is mediated by interaction between GST-S2-EGFP and the N-terminal 70 kDa aspect of FN. To determine whether there is a correlation between FN matrix levels and amount of associated S2 fusion protein, we used normal human lung fibroblasts (NHLFs), which substantially increase FN matrix assembly upon treatment with TGFß11. Cells were treated or not with TGFß during 3 days of growth to generate cultures with different levels of FN matrix. Cell lysates were collected and examined by immunoblotting (Supplementary Figure S4). To allow the calculation of the mass of each protein in the lysates, known masses of GST-S2-EGFP and FN were included on the immunoblots. Lysates from untreated cells contained approximately 3 nmol of FN and 0.16 nmol of GST-S2-EGFP, showing sub-stoichiometric levels of the S2 fusion protein relative to FN. TGFß induced a significant 5-fold increase in FN to ~ 16 nmol total. GST-S2-EGFP also increased in the TGFß lysates by 2.5-fold to approximately 0.4 nmol. Thus both proteins are increased with TGFß treatment, albeit not to the same proportion. GST-S2-EGFP stably colocalizes with FN fibrils in decellularized ECM After demonstrating that GST-S2-EGFP colocalizes with FN fibrils in growing cell cultures, we tested the stability of its association using decellularized matrices produced by NIH 3T3 fibroblasts. Cells were cultured under our standard conditions for decellularization supplemented with 2.5 µg/mL GST-S2-EGFP and were then decellularized with NP-40 detergent, low salt, high pH buffers23. After decellularization, matrices were stored in PBS at 4 ˚C and, after varying periods of time, GST-S2-EGFP association with FN fibrils was assessed. At 1, 4, 7, and 10 days after decellularization, matrices were solubilized in 2% SDS buffer and proteins were analyzed by immunoblotting with anti-FN and anti-GFP antibodies. Band intensities of FN were not statistically different across all time points (Fig. 5A), indicating that there is not significant loss of FN over the 10-day period. Similarly, there was little change in GST-S2-EGFP band intensities over time (Fig. 5B), demonstrating that the interactions between FN and GST-S2-EGFP are highly stable. Quantification of band intensities normalized to FN levels on day 1 show that there was substantially less GST-S2-EGFP than FN in the decellularized matrices. Fig. 5Quantification of GST-S2-EGFP and FN in cell-free matrices over time. NIH 3T3 fibroblasts were grown in the presence of 2.5 µg/mL GST-S2-EGFP for four days, then decellularized and stored in PBS. At 1, 4, 7, and 10 days post-decellularization, samples were solubilized in SDS buffer and equal volumes of each sample were analyzed on immunoblots for FN (A) and GST-S2-EGFP (B). Exposure time for FN was ~ 7 times shorter than for GST-S2-EGFP. Molecular weight markers (in kDa) are indicated on the left. These uncropped immunoblots are representative of 3 independent experiments. Quantification below each lane is band intensity (mean ± SD) normalized to day 1 FN for FN and GST-S2-EGFP bands. We found no significant differences between timepoints for either FN (p = 0.83) or GST-S2-EGFP (p = 0.16), as determined by one-way ANOVA. Quantification of GST-S2-EGFP and FN in cell-free matrices over time. NIH 3T3 fibroblasts were grown in the presence of 2.5 µg/mL GST-S2-EGFP for four days, then decellularized and stored in PBS. At 1, 4, 7, and 10 days post-decellularization, samples were solubilized in SDS buffer and equal volumes of each sample were analyzed on immunoblots for FN (A) and GST-S2-EGFP (B). Exposure time for FN was ~ 7 times shorter than for GST-S2-EGFP. Molecular weight markers (in kDa) are indicated on the left. These uncropped immunoblots are representative of 3 independent experiments. Quantification below each lane is band intensity (mean ± SD) normalized to day 1 FN for FN and GST-S2-EGFP bands. We found no significant differences between timepoints for either FN (p = 0.83) or GST-S2-EGFP (p = 0.16), as determined by one-way ANOVA. Parallel decellularized matrices were immunostained for FN and then images were captured for both FN and GST-S2-EGFP (Supplementary Figure S5). The FN matrix showed dense, fibrillar staining. The GST-S2-EGFP signal was also fibrillar and appeared to coincide with FN fibrils but was less dense than FN. We quantified and averaged mean fluorescence intensities (MFIs) for FN and GST-S2-EGFP to assess changes over time. Neither FN nor GST-S2-EGFP showed significant differences in average MFI throughout the 10-day storage in PBS, supporting the finding from the immunoblot analysis that the GST-S2-EGFP association with FN remains stable over time. While similar features can be seen between FN and GST-S2-EGFP in these images, the FN matrix was very dense and individual FN fibrils are difficult to make out. To allow us to assess colocalization between GST-S2-EGFP and FN fibrils in decellularized matrices, we developed a second decellularization procedure that uses a deoxycholate (DOC) buffer24. This procedure is mechanically gentler than our standard NP-40 procedure, thus allowing us to decellularize sparser matrices without significantly disrupting the fibril architecture. Decellularized ECMs prepared with the DOC procedure were examined by fluorescence microscopy at increasing times of storage. The calculated average MFIs for FN and for GST-S2-EGFP images showed that there was no significant difference in signals between timepoints (Fig. 6). Importantly, the reduced density of FN fibrils in these matrices makes the individual FN fibrils clearly distinguishable and the majority of fibrils are labeled with GST-S2-EGFP. Fig. 6GST-S2-EGFP colocalizes with FN fibrils through decellularization. Cells were cultured as in Fig. 5 except at lower cell density and with 20 µg/mL GST-S2-EGFP. Cultures were decellularized with a gentler, detergent-based procedure to preserve a sparser fibrillar network. At increasing times of storage in PBS, samples were fixed and immunostained for FN. Mean fluorescence intensities were quantified for 8 images in each of 3 samples and averages were normalized to Day 1 FN stain. Average values ± SD are listed on the images. We found no significant difference between FN levels (p = 0.86) or between GST-S2-EGFP levels (p = 0.66) across four timepoints as determined by one-way ANOVA. Scale bar = 200 μm. GST-S2-EGFP colocalizes with FN fibrils through decellularization. Cells were cultured as in Fig. 5 except at lower cell density and with 20 µg/mL GST-S2-EGFP. Cultures were decellularized with a gentler, detergent-based procedure to preserve a sparser fibrillar network. At increasing times of storage in PBS, samples were fixed and immunostained for FN. Mean fluorescence intensities were quantified for 8 images in each of 3 samples and averages were normalized to Day 1 FN stain. Average values ± SD are listed on the images. We found no significant difference between FN levels (p = 0.86) or between GST-S2-EGFP levels (p = 0.66) across four timepoints as determined by one-way ANOVA. Scale bar = 200 μm. GST-S2-EGFP tracks FN matrix assembly and fibril rearrangements in real-time Having established that the GST-S2-EGFP fusion protein binds stably to and colocalizes with FN fibrils, we sought to use this fusion protein to track assembly and rearrangement of FN matrix fibrils in real time. The S2 fusion protein and exogenous FN were added to cultures of NIH 3T3s and NHLFs that were actively assembling FN fibrils and then S2-labeled cultures were imaged for up to 20 h. Immediately after live imaging, cultures were fixed and immunostained for FN. Comparison of the GST-S2-EGFP signal from the final frame of the live imaging with the FN immunostain shows significant colocalization of fusion protein with FN fibrils for both mouse and human cultures (Fig. 7A). Thus the GST-S2-EGFP signal during live imaging faithfully depicts FN fibrils. Fig. 7GST-S2-EGFP enables live imaging of FN matrix assembly. Cells were grown for 3 days to allow initiation of matrix assembly. GST-S2-EGFP, FN, and Hoechst dye were added and, after up to 2 h, live imaging was initiated. Images were taken every hour for 18 h. (A) Images show the GFP signal from the final frame of GST-S2-EGFP live imaging (left) alongside post-imaging FN immunostaining (right) for the same field of view, for both NIH 3T3 and NHLF live imaging experiments. (B) Representative NIH 3T3 live imaging shows accumulation of GST-S2-EGFP-labeled FN fibrils over time. (C) GST-S2-EGFP-labeled small FN aggregates in an NIH 3T3 culture at 6 h (pink and yellow boxes) were observed to coalesce into fibrillar structures that then appear to merge at 12 h. (D) A GST-S2-EGFP-labeled FN aggregate is shown being stretched into a fibril between two cells with labeled nuclei (magenta). GST-S2-EGFP enables live imaging of FN matrix assembly. Cells were grown for 3 days to allow initiation of matrix assembly. GST-S2-EGFP, FN, and Hoechst dye were added and, after up to 2 h, live imaging was initiated. Images were taken every hour for 18 h. (A) Images show the GFP signal from the final frame of GST-S2-EGFP live imaging (left) alongside post-imaging FN immunostaining (right) for the same field of view, for both NIH 3T3 and NHLF live imaging experiments. (B) Representative NIH 3T3 live imaging shows accumulation of GST-S2-EGFP-labeled FN fibrils over time. (C) GST-S2-EGFP-labeled small FN aggregates in an NIH 3T3 culture at 6 h (pink and yellow boxes) were observed to coalesce into fibrillar structures that then appear to merge at 12 h. (D) A GST-S2-EGFP-labeled FN aggregate is shown being stretched into a fibril between two cells with labeled nuclei (magenta). Accumulation of GST-S2-EGFP-labeled FN fibrils was commonly observed in the videos. Frames over 16 h taken from Supplementary Video S1 clearly show lengthening, elongation and thickening over time of fibrils spanning multiple cells (Fig. 7B). Existing fibrils brightened over time as FN and S2 fusion protein were added to them, short fibrils merged to form longer ones, and new fibrils appeared in areas devoid of matrix. Areas of lower cell density tended to show fibril rearrangements that corresponded with cell movements including fibrils that moved into or out of the field of view. MFIs for first and last frames were calculated at six different locations from this sample and showed an average fold-increase in FN matrix of 3.22 ± 0.38. In terms of de novo fibrillogenesis, fibrils were observed to form through coalescence of small FN aggregates or puncta into a linear array (Fig. 7C; Supplementary Video S2). In some cases, coalescence was associated with the early stages of cell division (Supplementary Figure S6), perhaps to redistribute FN to opposite poles of the dividing cell. We also observed elongation of FN puncta. For example, Fig. 7D shows a small FN aggregate being stretched into a fibril by adjacent cells (Supplementary Video S3). Overall, our live imaging with the S2 fusion protein provides examples of FN fibril growth, remodeling and rearrangements, while also uncovering novel aspects of cell-matrix dynamics. Discussion Fibronectin (FN) matrix assembly plays a central role in tissue morphogenesis and disease and facilitates the deposition and organization of other ECM components. FN fibrillogenesis is a dynamic assembly process that depends on cell-FN and FN-FN interactions to yield a fibrillar network. Fully understanding this process requires visualizing assembly in live-cell systems. In this study, we identified a 12-amino acid peptide, S2, that enables fluorescent labeling of FN matrix in both live and fixed cell systems. The S2 peptide derived from an M13 phage display library panned against the 70 kDa N-terminal FN fragment outperformed other peptide candidates in promoting phage binding to matrix and incorporation into FN fibrils. Surprisingly, when displayed within the GST-S2-EGFP fusion protein, but not within the cleaved S2-EGFP construct, the peptide directs fusion protein incorporation into FN fibrils thus generating a fluorescent FN matrix. In cultures of both human and mouse cells, GST-S2-EGFP rapidly bound to preexisting FN matrix, was incorporated into newly assembled fibrils, and remained stably associated with insoluble FN matrix over time. Beyond enabling generation of fluorescent ECM, the S2 fusion protein permitted live imaging of ECM formation, revealing intermediate structures in the progression of FN fibril nucleation, elongation, accumulation, and remodeling over time. Given its versatility, S2-based proteins could be applied across cell culture systems to detect ECM alterations arising from diverse perturbations including drug treatment, mechanical stress, or disease-associated mutations. Live-cell imaging with GST-S2-EGFP revealed that FN matrix accumulation occurred preferentially in regions of high cell density, and that early deposition of nascent fibrils promoted subsequent regional matrix growth. At higher magnification, we observed cells consolidating punctate FN structures into fibrils and elongating globular aggregates into fibrillar forms. These observations showcase that FN fibrillogenesis can proceed through the maturation of pre-deposited intermediates, in which punctate or small globular deposits are remodeled into fibrils by cells. The reorganization of these pre-deposited structures into fibrillar networks indicates that early stages of FN matrix are not static but remain highly dynamic, a feature that is difficult to observe with fixed samples. Astrof and colleagues20 used single-molecule imaging to show fibrils composed of FN nanodomains in linear arrays. These nanodomains are significantly smaller than the FN puncta we observed, which likely represent larger nucleation structures for nascent fibrils, the physical properties and molecular composition of which are unexplored. It is possible that our small FN aggregates represent phase-like condensations of FN molecules at the cell surface, forming dynamic, liquid-like intermediates that subsequently stabilize into fibrils. Phase separation has been noted to be a driver of assembly of other ECM proteins25. The plasticity of early FN assembly emphasizes that FN matrix organization is a dynamic continuum rather than a binary state of ‘fibrils’ versus ‘non-fibrils.’ Such remodeling is likely critical for tissue homeostasis and could be dysregulated in disease contexts. We furthermore saw rearrangements of FN structures prior to and throughout cell division, highlighting how this fusion protein may uncover new aspects of cell–matrix dynamics. Interestingly, while the N-terminal location of the S2 peptide in the phage-displayed context was sufficient for matrix binding, N-terminal exposure of the S2 peptide in the fusion protein by removal of the GST moiety completely ablated binding to FN fibrils. Each phage displays 5 copies of the peptide-coat protein fusion clustered on its surface. It is possible that dimerization of GST26,27 may enhance fusion protein binding by presenting multiple S2 copies in close proximity, as in the phage. It is also possible that proteins flanking the peptide support conformational restrictions on the peptide sequence that promote stable FN interactions. Although the precise requirements remain unclear, both the S2-phage and the S2 fusion protein were readily incorporated into FN fibrils, suggesting the constraints on S2-mediated binding are not stringent. In terms of sequence homology, NCBI BLAST analysis of the S2 peptide found only short segments of 5 or fewer amino acids with homology to human proteins. The full 12-amino acid sequence does not exist natively in the human proteome, supporting the idea that S2 is a unique sequence that likely mediates a novel mode of FN recognition. GST-S2-EGFP offers several advantages for investigating the FN matrix. It is applicable to virtually all mammalian in vitro model systems without genetic manipulation, and labels endogenously-produced FN without further intervention. Unlike addition of exogenous FN, which can substantially affect matrix assembly, the S2 fusion protein does not change FN concentration. It is detectable both in live experiments and in fixed samples, both visually via fluorophores and biochemically. Its incorporation is robust and survives decellularization to generate fluorescent FN matrices, which may be useful tools in their own right. Others have identified FN binding peptides and used them to identify important properties of FN fibrils21,22. Although other FN-binding peptides have been described, none have been shown to drive the incorporation of a larger protein (~ 50 kDa in our case) into assembling FN fibrils or to persist within insoluble matrix through decellularization. Thus the S2 peptide reveals biochemical and functional properties not achieved by previously reported FN-binding reagents. These features allowed us to use the S2 peptide to integrate a fluorescent protein into FN matrix across species, which has already yielded insights about matrix assembly and supports the use of this peptide in other contexts. The modularity of the S2 fusion protein further expands its potential utility, allowing production of versions with different fluorescent moieties. Importantly, we have found that the S2 fusion protein can be added to without affecting matrix binding. Specifically, we fused the 27 kDa BioID2 enzyme onto the C-terminus of GST-S2-EGFP and successfully targeted the resulting ~ 75 kDa protein to FN matrix. This proof-of-principle construct illustrates that targeting a variety of functions to the FN matrix may be possible, such as alternative fluorescent proteins for multiplexed ECM imaging, enzymes to modify or define the ECM directly, or additional ECM proteins to alter matrix composition. FN is upregulated in many pathological contexts—including fibrosis and cancer—where it contributes to disease progression10,11,13,28. We found that TGFβ stimulation of human lung fibroblasts increased both FN and GST-S2-EGFP deposition. TGFβ is often used in cell culture models to simulate fibrotic disease states29 and is a well-known instigator of fibrosis in vivo30. Our results open the possibility that S2-guided molecules may also concentrate at sites of fibrosis in vivo, making such a molecule useful as a diagnostic imaging agent for FN-rich lesions or as a delivery platform for therapeutics targeting fibrotic or tumor-associated ECM. In summary, the S2 fusion protein forms specific, strong, and durable interactions with FN matrix fibrils. We believe it shows immediate promise as a versatile tool to investigate and to modify the ECM in experimental settings and shows potential as a therapeutic delivery system to FN-rich disease contexts as well. Methods Preparation of 70 kDa FN fragment Human 70 kDa cDNA was cloned into pMAX expression vector (Addgene) between a CMV promotor and SV40 poly(A) tail sequence. The pMAX backbone also contains a kanamycin resistance gene for selection. The pMAX-70 kDa plasmid was transfected into HEK293 cells using a Lipofectamine 3000 protocol (Invitrogen), transfectants were selected with G418 (ThermoFisher) and conditioned medium of clones was screened by immunoblotting with R457 rabbit antiserum raised in-house against 70 kDa protein. HEK293 cells were grown in Minimal Essential Medium plus 10% fetal bovine serum (Hyclone) and antibiotic/antimycotic cocktail. Human plasma FN was purified from fresh-frozen plasma (Innovative Research Inc) and 70 kDa protein was purified from HEK293 cell conditioned medium by gelatin-Sepharose affinity chromatography. Proteins were > 95% pure by Coomassie blue staining after SDS-PAGE and were stored at – 80˚C in 10 mM CAPS, pH 11, 150 mM NaCl (CAPS buffer). Phage display screening and ELISA The NEB PhD 12 library (New England Biolabs, NEB) was used with 70 kDa in a surface panning procedure following manufacturer instructions. A sterile polystyrene Petri dish was coated with a 25 µg/mL solution of 70 kDa in CAPS buffer overnight at 4˚C followed by blocking with BSA and washing with wash buffer as recommended. A 100-fold representation of the library (2 × 1011 pfu/mL) in wash buffer was incubated with the 70 kDa-coated dish for 20 min at room temperature with gentle rocking. After washing 10 times with wash buffer, bound phages were eluted with elution buffer (0.2 M glycine-HCl, pH 2.2, 1 mg/mL BSA) and neutralized with 1 M Tris-HCl, pH 9.1. Eluted phages were amplified with E. coli strain ER2738 and prepared for a second round of panning as recommended. Two subsequent rounds of panning were conducted. The final elution from the third round of panning was used to infect E. coli without amplification and 50 individual phage plaques were selected and amplified for DNA sequencing using provided primers flanking the insert region. ELISAs were conducted as recommended by NEB. Amplified and concentrated phage stocks were prepared and 100 µL of 1011 pfu/mL in TBS-Tween was incubated with 70 kDa-coated wells in flat bottom, 96-well plates. Bound phages were detected with HRP-conjugated anti-M13 antibody (GE Healthcare) and ABTS plus H2O2 and measured at 405 nm wavelength by BioRad iMark microplate reader. Matrix binding and incorporation assays using phage NIH 3T3 mouse fibroblasts (ATCC CRL-1658) were seeded in DMEM (high glucose) plus 5% bovine calf serum (Hyclone) at 50,000 cells per cm2 on glass coverslips and allowed to grow for 48 h to establish a dense ECM. Monolayers were then washed with PBS, fixed with 3.7% formaldehyde in PBS, blocked with 1 mg/mL BSA in PBS for 30 min, and incubated with phage at 2 × 1012 pfu/mL in PBS for 30 min. Unbound phage were washed away with PBS + 0.1% Tween-20, and samples were fixed as above and stained for FN (rabbit R184 antiserum raised in-house against FN III1 − 631) and M13 phage (Sino Biological). Images were captured using Nikon Eclipse Ti microscope equipped with a Hamamatsu C11440 ORCA-Flash 4.0 LT plus digital camera and NIS-Elements Software (Nikon). Images compared with one another were equivalently adjusted for brightness and contrast in FIJI for presentation in figures. To characterize phage binding in this assay, representative images from each phage were chosen and separated into FN, phage, and merge channels. Channels were examined side by side and positive binding was attributed to the phage if the pattern of the phage signal resembled the fibrillar structure of the matrix, and the merge channel depicted colocalization of the majority of the phage signal with FN fibrils. For phage incorporation during matrix assembly, NIH 3T3 fibroblasts or human WI-38 fibroblasts were seeded on glass coverslips as above. WI-38 human diploid fibroblasts (ATCC CCL-75) were grown in Modified Eagle’s Medium containing 1% nonessential amino acids and 1mM sodium pyruvate (Life Technologies) with 10% fetal bovine serum (Hyclone), 4 mM glutamine (ThermoFisher) and penicillin/streptomycin (ThermoFisher). Following cell attachment, phages were added at 2 × 1012 pfu/mL, and cells were grown for 48 h to confluence. Coverslips were immunostained and visualized as above. Pearson’s correlation coefficients were generated using the FIJI JACoP plugin. Fusion protein generation and purification Peptide sequences were inserted between GST and EGFP sequences in pGEX4T3-GST-EGFP expression plasmid (Addgene) using the Q5® Site-Directed Mutagenesis Kit (NEB). The manufacturer’s instructions were used to design the primers (F: 5’ –GCCGGGTCTCGAGGGTGGTGGTATGGTGAGCAA GGGCGAG – 3’; R: 5’ – AGTTTGGTCCAGTCATAGCTCCCGCCCTGAAAATACAGGTTTTCG – 3’), perform the PCR amplification, and carry out kinase, ligase & DpnI treatment. The ligated product was transformed into DH5⍺ competent E. coli (NEB), DNA was extracted from positive clones (QIAprep Spin Miniprep Kit, Qiagen) and the fusion protein-peptide sequence was confirmed by whole plasmid nanopore sequencing (Plasmidsaurus Inc.). The resulting GST-peptide-EGFP plasmids were transformed into BL21 competent E. coli (NEB) for protein expression and purification. EGFP was replaced with an mScarlet cDNA amplified using the pmScarlet_C1 plasmid (Addgene) with the addition of XhoI and BsrGI restriction sites. The GST-S2-EGFP plasmid and mScarlet amplicon were digested using XhoI and BsrGI, mixed, ligated, and transformed into competent DH5⍺ E. coli. Subsequent steps were as described above. E. coli BL21 (DE3) cells expressing GST-S2-EGFP, GST-S2-mScarlet, GST-control-EGFP and GST-S3-EGFP were used for protein expression. The control is a 6-mer peptide with sequence SNSRVD. Proteins were purified from IPTG-treated cultures by glutathione-Sepharose 4B affinity chromatography in cell disruptor fluid (50 mM Tris-HCl, 50 mM NaCl, 2 mM EDTA, pH 8.0) and, after elution with buffer plus glutathione, peak fractions were pooled and dialyzed into cell disruptor fluid at pH 7.5. Purity was monitored by Coomassie blue staining after SDS-PAGE and proteins were stored at -80°C. GST tag removal was performed according to the NEB protocol. Protein bound to glutathione resin in 50 mM HEPES pH 7.4, 100 mM NaCl was incubated with His-tagged TEV protease (NEB) overnight at 4 °C with rotation. Supernatant containing S2-EGFP and TEV protease was then incubated with Ni-NTA agarose resin (Thermo-Fisher) for 1 h at 4 °C to remove the His-tagged TEV protease. Purity of cleaved S2-EGFP protein was confirmed by SDS-PAGE. Protein was stored at − 80 °C. The presence of the peptide in the fusion proteins before and after TEV cleavage was confirmed by mass spectrometry (Princeton Molecular Biology Proteomics Facility). S2 fusion protein incorporation into FN matrix For matrix labeling with GST-S2-EGFP, NIH 3T3 fibroblasts were grown as for phage matrix incorporation but with the addition of fusion proteins at 2.5 to 20 µg/ml as indicated. Cells were incubated for 48 h and innate fluorescence of the EGFP proteins was used to visualize the fusion protein. FN fibrils were detected after fixation by staining with R184 anti-FN antiserum as described above. In some assays, human FN was added at 10 µg/mL along with the fusion protein and visualized with an anti-human FN monoclonal antibody (HFN 7.1, DSHB). Normal human lung fibroblasts (NHLFs; Lonza Biosciences) were cultured in FBM-2 fibroblast growth medium-2 bulletkit according to manufacturer’s instructions. NHLFs were grown with 2.5 µg/ml GST-S2-EGFP and with or without the addition of 5 ng/ml TGF-ß1 (Peprotech) as described11. Matrix decellularization Decellularized ECM was produced using two different methods. In the first method, NIH 3T3 cells were cultured in medium supplemented with 2.5 µg/mL GST-S2-EGFP and, after 4 days of growth, cultures were decellularized using a low salt/high pH lysis buffer (8 mM sodium phosphate pH 9.6, 1% NP-40)23,32. After final washes, matrices were stored in PBS at 4 ˚C. The second decellularization procedure, which requires fewer buffer exchanges and washing steps, was adapted from the method of Quade and McDonald24. In brief, cells were washed in PBS, then permeabilized for 3 min using 3% Triton-X100 in PBS followed by treatment with 20 U/mL DNase I (NEB) for 10 min. DNase solution was aspirated and samples were treated with lysis buffer (2% DOC, 20 mM Tris pH 8.8, 2.5 mM EDTA) for 15 min at 37 ˚C, followed by an additional 10 min in fresh lysis buffer. Decellularized matrices were washed twice with PBS and stored in PBS at 4˚C until use. For immunofluorescence analyses, decellularized matrices were fixed as above and stained with R184 anti-FN antiserum and AlexaFluor 568-conjugated goat anti-rabbit IgG (1:500) (Invitrogen). Samples were mounted using Vectashield antifade mounting medium (Invitrogen). Images were captured using a Nikon Eclipse Ti microscope as described above. Images within each figure were equivalently adjusted for brightness and contrast using FIJI. Immunoblotting Decellularized matrices were solubilized in SDS buffer (2% SDS, 100 mM Tris-HCl, pH 8.8, 2 mM EDTA) by shearing through a 26G needle and boiling. SDS-solubilized decellularized material was separated by SDS-PAGE using 0.5 µl of material for FN and 2.0 µl for GST-S2-EGFP immunoblots. NHLF cultures were solubilized in 2% deoxycholate (DOC), 50 mM Tris-HCl, pH 8.8, 2 mM EDTA and, after shearing, insoluble material was removed by centrifugation. Total protein concentrations were determined with a BCA assay (Pierce) and equal amounts of total protein were separated by SDS-PAGE on 10% (for GST-S2-EGFP) or 6% (for FN) polyacrylamide gels. Gels were transferred to nitrocellulose membrane and immunoblotted as previously described33. Primary antibodies used are R184 anti-FN at 1:2000 dilution, and mouse monoclonal anti-GFP (ABclonal) at 1:2500. Secondary HRP-conjugated anti-rabbit and anti-mouse antibodies (ThermoFisher Scientific) at 1:10,000 dilution followed by SuperSignal West Pico Plus ECL reagent (Pierce) were used for developing blots. An Azure 600 imaging system (Azure Biosystems) was used for imaging blots, and AzureSpot Pro was used for band density analysis. Quantification was taken from band intensities within the linear range. Confocal microscopy and live imaging High-resolution imaging of FN matrix and phage was performed using a Nikon A1R-Si HD confocal microscope (Nikon Instruments). Images were acquired with a 60x oil-immersion objective to visualize FN fibrils within the ECM. To assess fine structural details, 100x oil-immersion objective was applied to capture individual fibrils. Laser excitation and detector settings were optimized to minimize photobleaching while maintaining maximal signal-to-noise ratio. All image acquisition parameters were kept consistent across experimental conditions to allow for direct comparison. Images were denoised using NIS Elements software version 4.11.0. For live imaging experiments, NIH 3T3 cells were seeded at 0.8 × 104 cells/cm2 and NHLFs were seeded at 1.3 × 104 cells/cm2 on glass-bottom culture dishes (Mattek). In experiments shown in Figs. 7A, B and Supplementary Video 1, the glass-bottom surface was precoated with 10 µg/ml bovine collagen I solution (Sigma). After 3–4 days of growth, exogenous FN and GST- S2-EGFP were added at 20 µg/ml each and Hoechst dye was added at 5 µg/ml. 20 to 120 min later, cultures were imaged using a 20X objective on an inverted Nikon Eclipse Ti2 spinning disk confocal microscope equipped with a Tokai Hit stage-top incubator, maintained at 37 °C with 5% CO2. Cells were then imaged hourly for up to 20 h at several locations for Hoechst dye and GST-S2-EGFP using the 405 nm laser (15% intensity, 400 msec exposure) and 488 nm laser (6% intensity, 500 msec exposure time), respectively. Brightfield images were taken of each location before and after live imaging. In some experiments, the monolayers were fixed and immunostained for FN immediately afterward and images were captured using a Nikon Eclipse Ti as described above to assess colocalization between GST-S2-EGFP and FN. Images were denoised using NIS Elements software version 5.3. Statistical analyses Mean fluorescence intensity (MFI) for each image was quantified using FIJI, and Pearson’s correlation coefficients were calculated using the FIJI JaCoP plugin. MFIs and immunoblot quantifications are reported as mean ± standard deviation. Statistical tests used were a two-tailed t test for a difference between means or one-way ANOVA. p values ≤ 0.05 were considered statistically significant. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4