Collagen-I is endocytosed and reassembled into fibrils

A. Fluorescent images of tail tendon incubated with Cy3-colI for 5 days, showing presence of collagen-I within the cells, and fibril-like fluorescence signals outside of cells. Hoechst stain was used to locate cells within the tendon. Area surrounded by yellow box expanded on the right, and cells with Cy3-colI present intracellularly pointed out by yellow triangles. Area surrounded by grey box expanded on the right, and fibril-like fluorescence signals indicated with white arrows. Scale bar = 50 µm. Representative of N = 3.

B. Bar chart showing an increase of percentage of fluorescent iTTFs incubated with 1.5 µg/mL Cy3-colI over time (left), and an increase of percentage of fluorescent iTTFs incubated with increasing concentration of Cy3-colI for one hour (right), suggesting a non-linear time-dependent and dose-dependent uptake pattern. N = 3.

C. Flow cytometry imaging of iTTFs incubated with 5 µg/mL Cy3-labeled collagen-I for one hour, showing that collagen-I is taken up by cells and held in vesicular-like structures. Images acquired using ImageStream at 40x magnification. Scale bar = 10 μm. Cy3 – Cy3 channel, BF/Cy3 – merged image of BF and Cy3. Representative of >500 cells images collected per condition.

D. Fluorescent images of iTTFs transduced with Rab5-GFP and incubated with Cy3-labeled collagen-I. Yellow arrows point to labelled collagen co-localizing with Rab5 in intracellular structures. Representative of N = 3. Scale bar = 10 µm.

E. Fluorescent images of iTTFs incubated with 5 µg/mL Cy3-colI for one hour, trypsinized and replated in fresh media, and further cultured for 72 h. Top labels denotes the fluorescence channel corresponding to proteins detected. Merged image colour channels as denoted on top left. Representative of N>3. Scale bar = 20 µm.

F. Fluorescent image series of Cy3-colI incubated at different concentrations for 72 h, either cell-free (right panel), or with iTTFs (+fibroblasts, left panel). Representative of N = 3. Scale bar = 20 µm. Red box - zoomed out to the bottom left and separated according to fluorescence channel. White arrows highlighting Cy3-positive fibrils assembled by fibroblasts when incubated with 0.1 µg/mL Cy3-colI.

G. Quantification of the area of Cy3-positive fibrils in cell-free cultures, quantified per image area. N=3.

H. Quantification of the area of Cy3-positive fibrils in +fibroblasts cultures, corrected to number of nuclei per image area. N=3.

I. Comparison of total area of Cy3-positive fibrils in cell-free and +fibroblast cultures at 0.1 µg/mL concentration, as quantified per image area. N=3. ****p <0.0001.

Inhibition of endocytosis leads to changes in collagen-I homeostasis, and endocytosis is a rhythmic event

A. Left: fluorescent images of collagen-I (red) counterstained with DAPI (blue) in iTTFs treated with DMSO (top) or Dyng4a (bottom) for 72 h. Scale bar = 20 µm. Right: quantification of area occupied by collagen-I fibrils, corrected to number of nuclei. N = 3 with 5 images from each experiment ** p = 0.0025.

B. Western blot analysis of conditioned media taken from iTTFs treated with DMSO or Dyng4a for 72 h, showing a decrease in collagen-I secretion. Top: probed with collagen-I antibody (Col-I), bottom: counterstained with Ponceau (Pon) as control. Protein molecular weight ladders to the left (in kDa). Representative of N = 3.

C. Left: fluorescent images of fibronectin (magenta) counterstained with DAPI (blue) in iTTFs treated with DMSO (top) or Dyng4a (bottom) for 72 h. Scale bar = 20 µm. Right: quantification of area occupied by fibronectin fibrils, corrected to number of nuclei. N = 3 with 5 images from each experiment.

D. Percentage Cy3-colI taken up by synchronized iTTFs over 48 h. Meta2d analysis indicates a circadian rhythm of periodicity of 23.8 h. Bars show mean ± s.e.m. of N = 3 per time point.

E. Percentage of Cy3-colI taken up by synchronized iTTFs, corrected to the maximum percentage uptake of the time course (pink, bars show mean ± s.e.m. of N=3 per time point), compared to the percentage collagen fibril count over time, corrected to the maximum percentage fibril count of the time course (black, fibrils scored by two independent investigators. Bars show mean± s.e.m. of N=2 with n = 6 repeats at each time point).

Collagen-I recycling can generate fibrils

A. Fluorescent image series of iTTFs treated with scrambled control (top panel, scr), and siRNA against col1a1 (bottom panel, siCol1a1). Labels on top denotes the fluorescence channel corresponding to proteins detected (ColI – collagen-I Fn1 – fibronectin);. Quantification of collagen-I and fibronectin signal to the right. Representative of N = 4. Scale bar = 25 µm. * p = 0.021.

B. Fluorescent image series of scr (left column) and siCol1a1 (right column) iTTFs incubated with Cy3-colI. Labels on left denotes the fluorescence channel(s) corresponding to proteins detected (ColI – collagen-I Fn1 – fibronectin). Cy3-colI fibrils highlighted by red arrows, and collagen-I fibrils highlighted by white arrows. Both scr cells and siCol1a1 cells can take up exogenous collagen-I and recycle to form collagen-I fibril. Representative of N>3. Scale bar = 10 μm.

C. Fluorescent image series of siCol1a1 iTTFs treated with DMSO control (left) or Dyngo4a (right) during Cy3-colI uptake, followed by further culture for 72 h. Labels on left denotes the fluorescence channel corresponding to proteins detected (ColI – collagen-I Fn1 – fibronectin). Quantification of Cy3-colI signal to the bottom. Dyngo4a treatment led to a reduction of Cy3-colI fibrils. Representative of N>3. Scale bar = 20 µm. ** p = 0.0022

Fibroblasts without endogenous collagen-I can effectively make fibrils by endocytic-recycling of exogenous collagen.

A. Fluorescent images of primary tail tendon fibroblasts isolated from control mice (top panel, CKO-), and tamoxifen-treated collagen-knockout mice (bottom panel, CKO+). Labels on top denotes the fluorescence channel corresponding to proteins detected. Quantification of collagen-I and fibronectin fluorescence signal to the right. Representative of N=3. Scale bar = 10μm. ** p = 0.0084.

B. Fluorescent images of CKO-/CKO+ tail tendon fibroblasts incubated with Cy3-colI. Labels on top denotes the fluorescence channel corresponding to proteins detected. Cy3-colI fibril highlighted by red arrows, and collagen-I fibril highlighted by white arrows. Both CKO- and CKO+ cells can take up exogenous collagen-I and recycle to form collagen-I fibrils. Representative of N>3. Scale bar = 25μm.

C. Fluorescent image series of CKO+ tail tendon fibroblasts treated with DMSO control (left) or Dyngo4a (right) during Cy3-colI uptake, followed by further culture for 72 h. Labels on left denotes the fluorescence channel corresponding to proteins detected. Quantification of Cy3-colI signal to the bottom. Dyngo4a treatment led to a significant reduction of Cy3-colI fibrils. Representative of N>3. Scale bar = 25 µm. * p = 0.00273.

VPS33B controls collagen fibril formation at the plasma membrane in a rhythmic manner

A. Electron microcopy images of fibroblasts plated on Aclar and grown for a week before fixation and imaging. Ctrl culture has numerous collagen-I fibrils, as pointed out by arrows. Yellow arrow points to a fibripositor, and green box is expanded to the left bottom corner, showing the distinct D-banding pattern of collagen-I fibril when observed with electron microscopy. VPSko clones all have fewer and thinner fibrils present in the culture (pointed out by red arrows). Representative of N = 3. Scale bar = 0.5 µm.

B. Fluorescence images of collagen-I (red) and DAPI counterstain in ctrl and VPSko iTTFs. Yellow arrows indicating collagen fibrils, and white arrows pointing to collagen-I presence in intracellular vesicles. Representative of N>6. Scale bar = 25 µm.

C. Matrix deposition by ctrl or VPSko iTTFs, after one week of culture. Left: decellularized matrix mass. N = 4, * p = 0.0299. Right: hydroxyproline content presented as a ratio between ctrl and VPSko cells. N = 4, *p = 0.0254. Ratio-paired t-test used.

D. Fluorescence images of collagen-I (red) and DAPI counterstain in ctrl and VPSoe iTTFs. Representative of N>6. Scale bar = 20 µm.

E. Matrix deposition by ctrl or VPSoe iTTFs, after one week of culture. Left: decellularized matrix mass, N = 4. Right: hydroxyproline content presented as a ratio between ctrl and VPSoe cells, N = 4. Ratio-paired t-test used.

F. Relative collagen fibril count in synchronized ctrl (black) and VPSoe (pink) iTTFs, corrected to the number of fibrils in ctrl cultures at start of time course. Fibrils scored by two independent investigators. Bars show mean± s.e.m. of N = 2 with n=6 at each time point.

G. Western blot analysis of conditioned media taken from ctrl and VPSoe iTTFs after 72 h in culture. Top: probed with collage-I antibody (ColI), bottom: counterstained with Ponceau (Pon) as control. Protein molecular weight ladders to the left (in kDa). Representative of N = 3.

Procollagen-I and VPS33B localize to the same compartments

A. Schematic depicting the proposed membrane topologies of VPS33b.

B. iTTFs expressing BFP-tagged VPS33B. Left: BFP tagged on the N-terminal end of VPS33B (VPSnBFP), Right: BFP tagged on the C-terminal end (VPScBFP). Images taken in Airy mode. Representative of N>4. Scale bar = 10 µm.

C. Schematic of the split-GFP system. GFP1-10 barrel is introduced into VPS33B (VPS-barrel), and GFP11 to alpha-1 chain of Collagen-I (GFP11-proα1(I)). If the two tagged proteins co-localize (e.g. in a vesicle), a GFP signal will be emitted.

D. Brightfield (top) and fluorescence (middle) images of iTTFs expressing both VPS-barrel and GFP11-proα1(I) constructs. Representative of N=5. Green box is expanded to the bottom, to highlight the punctate fluorescence signals within intracellular vesicular structures, as well as fibril-like structures suggestive of fibril assembly sites. Scale bar = 20 µm.

E. Fluorescence images of VIPAS (green), collagen-I (red) and DAPI counterstain in iTTFs. Representative of N=3. Green box is expanded to the right (flipped 90°) to show strong VIPAS signal encasing collagen-I. Scale bar = 25 µm.

F. Quantification of average number of fibrils per cell (left) and average fibril length (right) in control endogenously-tagged Dendra-colI expressing 3T3 cells (ctrl) and Dendra-colI expressing 3T3 overexpressing VPScBFP (VPScBFP). >500 cells quantified per condition. N=12. *P=0.048.

G. Brightfield (left) and fluorescence (middle) images of iTTFs expressing VPS-barrel incubated with conditioned media containing Col1a1-GFP11 for 24hr. Scale bar = 25μm.

H. Line charts comparing the percentage of iTTFs that have taken up 5 µg/mL Cy3-colI (left) and Cy5-colI (right) after one hour incubation between control (ctrl), VPS33B-knockout (VPSko) and VPS33B-overexpressing (VPSoe) cells, corrected to control. RM one-way ANOVA analysis was performed. N = 4.

I. Fluorescence images of iTTFs of different levels of VPS33B expression, fed with Cy5-colI and further cultured for 72 h. Cultures were counterstained with DAPI. Box expanded to right of images to show zoomed-in images of the fibrils produced by the fibroblasts. Representative of N = 2.

Integrin α11 subunit mediates VPS33B-effects and is required for collagen-I fibrillogenesis

A. Top 25 Functional Annotation of proteins detected in biotin-enriched samples when compared to non-enriched samples based on p-values. Y-axis denotes the GO term, X-axis denotes –log (P value).

B. Heatmap representation of spectral counting of collagens detected in biotin-enriched surface proteins from control (ctrl), VPS33B-knockout (VPSko), and VPS33B-overexpressing (VPSoe) iTTFs. Scale denotes quantitative value as normalized to total spectra, as determined by Proteome Discoverer.

C. Heatmap representation of spectral counting of integrins detected in biotin-enriched surface proteins from control (ctrl), VPS33B-knockout (VPSko), and VPS33B-overexpressing (VPSoe) iTTFs. Scale denotes quantitative value as normalized to total spectra, as determined by Proteome Discoverer.

D. Heatmap representation of spectral counting of Plod3 and VPS33B detected in biotin-enriched surface proteins from control (ctrl), VPS33B-knockout (VPSko), and VPS33B-overexpressing (VPSoe) iTTFs. Scale denotes quantitative value as normalized to total spectra, as determined by Proteome Discoverer.

E. Western blot analysis of integrin α11 subunit levels in control (ctrl), VPS33B-overexpressing (VPSoe), VPS33B-knockout (VPSko) iTTFs. Top: probed with integrin α11 antibody, bottom: reprobed with GAPDH antibody. Protein molecular weight ladders to the left (in kDa). Representative of N = 3.

F. qPCR analysis of Itga11 transcript levels in ctrl compared to VPSko iTTFs (left), and ctrl compared to VPSoe iTTFs (right). N>3, ****P<0.0001, *P = 0.0226.

G. qPCR analysis of Itga11 mRNA expression in ctrl (left) or VPSoe (right) iTTFs treated either with scrambled control (scr) or siRNA against Itga11 (siItga11), collected after 96 h. N=3, **P=0.0091, ****P<0.0001.

H. IF images of ctrl and VPSoe iTTFs treated either with control siRNA (scr) or siRNA again Itga11 (siItga11), after 72 h incubation; collagen-I (red) and DAPI (blue) counterstained. Representative of N = 3. Scale bar = 25 µm.

I. Bar chart comparing the percentage of iTTFs that have taken up 5 µg/mL Cy5-colI after one hour incubation between fibroblasts treated with scrambled control (ctrl) or siRNA against Itga11 (siItga11), corrected to scr. N=3. **P=0.0062.

Fibroblasts derived from IPF patients have higher collagen endocytic-recycling capacity that is mediated by VPS33B and ITGA11.

A. qPCR analysis of patient-derived fibroblasts isolated from control (ctrl) or IPF lungs. Bars showing mean± s.e.m, 5 patients in each group from 2 independent experiments (technical repeats not shown here). Itga11, *p = 0.0259; VPS33B, *p = 0.0183.

B. Fold change of percentage Cy5-colI (left) or Cy3-colI (right) taken up by ctrl or IPF lung fibroblasts, corrected to average of control fibroblasts. Bars showing mean± s.e.m., 5 patients in each group from 2 independent experiments (technical repeats not shown here). **p = 0.003.

C. Fluorescent images of ctrl or IPF lung fibroblasts that have taken up Cy3-colI (magenta), followed by further culture for 48 h in presence of ascorbic acid, before subjected to collagen-I staining (green). Labels on top denotes the fluorescence channel corresponding to proteins detected. Quantification of Cy3-colI signal to the right. IPF fibroblasts produced more Cy3-labelled fibrillar structures. Representative of N = 5. Scale bar = 20 µm. * p = 0.0135.

D. Fluorescent images of IPF lung fibroblasts treated with siRNA scrambled control (scr), siRNA against VPS33B (siVPS33B) or siRNA against ITGA11 (siITGA11) prior to uptake of Cy5-colI (magenta). This was followed by further culture for 48 h in presence of ascorbic acid, before subjected to collagen-I staining (green). Labels on top denotes the fluorescence channel corresponding to proteins detected. Quantification of Cy5-colI signal to the right. Both siVPS33B and siITGA11 significantly reduced recycled collagen signals. Representative of n = 5 across N = 2. Ordinary one-way ANOVA with multiple comparisions (to scr) was performed on quantification of Cy5-colI signal. siVPS33B, * p = 0.0341; siITGA11, * p = 0.0282.

The IPF fibrotic focus is positive for integrin α11 subunit and VPS33B.

A. Immunohistochemistry of IPF patient (patient 1) with red dotted line outlining the fibroblastic focus, the hallmark lesion of IPF. Sections were stained with hematoxylin and eosin (H&E), collagen-I (ColI), integrin-α11, VPS33B. Scale bar = 50 µm.

B. Immunohistochemistry of IPF patient 4 showing regions of emerging fibrotic remodeling with evidence of fibroblastic foci formation (red asterisks). Sections were stained with H&E, ColI, integrin-α11, VPS33B. Scale bar = 100 µm.

C. Immunohistochemistry of 5 µm thick sequential lung sections taken from lungs classified as control (Control 1). Sections were stained with hematoxylin and eosin (H&E), collagen-I (ColI), integrin-α11, VPS33B. Scale bar = 100 µm.

Proteins responsible for collagen fibrillogenesis are also co-localized to diseased areas of chronic skin wounds.

A. Immunohistochemistry of 5 µm thick sequential skin sections taken from normal skin regions of patients with chronic skin wounds (Patient 1, Patient 2, Patient 3, Patient 4). Sections were stained with pentachrome, integrin-α11, VPS33B. Scale bars positioned in top left corner: black (unzoomed pentachrome) = 100 µm, white (zoomed sections) = 50 µm.

B. Immunohistochemistry of 5 µm thick sequential skin sections taken from the chronic wound areas from patients with chronic skin wounds (Patient 1, Patient 2, Patient 3, Patient 4). Sections were stained with pentachrome, integrin-α11, VPS33B. Scale bars positioned in top left corner: black (unzoomed pentachrome) = 100 µm, white (zoomed sections) = 50 µm.

Proposed working model of collagen homeostasis in fibroblasts.

Endogenous collagen is either secreted as protomers (soluble secretion route, not circadian rhythmic) or made into fibrils (fibril assembly route, circadian rhythmic). Secreted collagen protomers can be captured by cells through endocytosis (circadian rhythmic) and recycled to make new fibrils. Integrin α11 and VPS33b directs collagen to fibril formation.