Fortilin interacts with TGF-β1 and prevents TGF-β receptor activation

Fortilin is a 172-amino acid multifunctional protein present in both intra- and extracellular spaces. Although fortilin binds and regulates various cellular proteins, the biological role of extracellular fortilin remains unknown. Here we report that fortilin specifically interacts with TGF-β1 and prevents it from activating the TGF-β1 signaling pathway. In a standard immunoprecipitation-western blot assay, fortilin co-immunoprecipitates TGF-β1 and its isoforms. The modified ELISA assay shows that TGF-β1 remains complexed with fortilin in human serum. Both bio-layer interferometry and surface plasmon resonance (SPR) reveal that fortilin directly bind TGF-β1. The SPR analysis also reveals that fortilin and the TGF-β receptor II (TGFβRII) compete for TGF-β1. Both luciferase and secreted alkaline phosphatase reporter assays show that fortilin prevents TGF-β1 from activating Smad3 binding to Smad-binding element. Fortilin inhibits the phosphorylation of Smad3 in both quantitative western blot assays and ELISA. Finally, fortilin inhibits TGFβ-1-induced differentiation of C3H10T1/2 mesenchymal progenitor cells to smooth muscle cells. A computer-assisted virtual docking reveals that fortilin occupies the pocket of TGF-β1 that is normally occupied by TGFβRII and that TGF-β1 can bind either fortilin or TGFβRII at any given time. These data support the role of extracellular fortilin as a negative regulator of the TGF-β1 signaling pathway.

O riginally cloned in 1988 by Gross et al 1 . as a cytosolic molecule abundantly expressed in tumor cells, fortilin remained a protein of unknown function until 2001 when we 2 and others 3,4 reported that it protects against apoptosis. Investigation over the next 20 years revealed that fortilin is a multi-functional protein implicated in diverse biological processes, including protection against apoptosis [2][3][4] , endoplasmic reticulum stress handling 5 , cell cycle progression 6 , reactive oxygen species detoxification 7 , and Ig-E-mediated histamine release 8,9 . Fortilin exists in both the cytosol and the nucleus (cellular fortilin) 2 and circulates in the blood 10 after being secreted from the cell (circulating fortilin) 11 .
In the cell, fortilin exerts its biological activities through its molecular interaction with its "executioner" proteins. For example, fortilin binds and negatively regulates p53 by preventing p53 from transcriptionally activating its target molecules BAX, PUMA, and NOXA 12,13 . In addition, fortilin binds and positively regulates peroxiredoxin-1, a reactive oxygen species-detoxifying enzyme, by preventing it from being inactivated by Mst1 kinase 7 . Moreover, fortilin binds and stabilizes the myeloid cell leukemia protein-1 (MCL1), a Bcl-2 family member anti-apoptotic molecule, thereby supporting the survival of myeloid cells 14,15 . Finally, fortilin binds and inhibits the inositol-requiring enzyme 1 alpha (IRE1α), an endoplasmic reticulum (ER) stress sensor 5 .
We and others have shown that fortilin is secreted from the cell 10 via the non-classical pathway 11 and circulates in the blood of humans and mice 10 . However, the biological activity of circulating fortilin has been poorly understood. Herein we report that fortilin physically binds TGF-β1, functionally inhibits the TGF-β1 signaling pathway, and prevents mesenchymal progenitor cells from differentiating into vascular smooth muscle cells. We propose that extracellular fortilin is an inhibitor of TGF-β1.

Results
Co-immunoprecipitation (Co-IP) assays showed that fortilin binds TGF-β1. To test whether fortilin physically interacts with TGF-β1, we performed in vitro Co-IP assays using recombinant human fortilin (rh-fortilin) and TGF-β1 (rh-TGF-β1) proteins. We equally divided the reaction mixture containing rh-fortilin, rh-TGF-β1, rh-p53, and rh-N-Ribosyldihydronicotinamide:Quinone Reductase 2 (rh-NQO2) into two microfuge tubes (  Fig. S4 hereafter). We added rabbit IgG to the first tube and rabbit monoclonal anti-fortilin antibody (α-fortilin mAb) to the second tube. After incubation and extensive washing, the immune complex pulled down with anti-rabbit-IgG magnetic beads was eluted into SDS-loading buffer. The system was functioning appropriately, as fortilin was able to co-immunoprecipitate p53, a known fortilin-interacting protein 12 (Fig. 1a, lane 3, row c). Washing was sufficiently stringent to prevent fortilin from nonspecifically binding NQO2, a protein that is known not to interact with fortilin 7 (Fig. 1a, lane 3, row d). Using this protocol, the presence of coimmunoprecipitated TGF-β1 was evaluated by the same western blot analysis. We found that α-fortilin mAb (Fig. 1a, lane 3, row a), but not IgG (Fig. 1a, lane 2, row a), immunoprecipitated fortilin and that TGF-β1 was successfully co-immunoprecipitated in the presence of fortilin (Fig. 1a, lane 3, row b) but not in its absence (Fig. 1a, lane 2, row b), suggesting that fortilin and TGF-β1 specifically interact with each other.
Fortilin circulates in the mouse and human blood at the serum concentrations of 2.17 ± 0.58 nM (biological replicates (N) = 30) and 3.41 ± 2.07 nM (N = 63), respectively, as we previously reported 10 . To assess the relative contribution of TGF-β1, TGF-β2, and TGF-β3 to the formation of a complex with fortilin in the blood, we measured their serum concentrations using a multiplex assay system (MILLIPLEX™ MAP TGFβ Magnetic Bead 3 Plex Kit, Millipore Sigma, Burlington, MA). We found the concentrations of TGF-β1, TGF-β2, and TGF-β3 to be 10.5 ± 2.1, 1.2 ± 0.06, and <0.023 (lower than the detection limit) nM, respectively (N = 15 each) (Fig. S1a), suggesting that the fortilin-TGF-β1 interaction represents the majority of fortilin-TGF-β interactions in the blood. Similar results have been reported for normal human plasma as well 16,17 . We therefore focused on the fortilin-TGF-β1 interaction in the remainder of the project.
Bio-layer interferometry (BLI) showed that fortilin binds TGF-β1. The above Co-IP, ELISA, and column experiments suggested that fortilin specifically interacts with TGF-β1 (Figs. 1, S1c). Next, we investigated binding between the two proteins using both BLI and surface plasmon resonance (SPR). For BLI assays, biotinylated rh-fortilin was immobilized on streptavidin-coated biosensors (ForteBio, Menlo Park, CA) and washed in PBS for 30 s. Various concentrations of rh-TGF-β1 were applied to the biosensors for 180 s to evaluate the association between the two proteins before rh-TGF-β1 solution was replaced by PBS for 180 s to evaluate their dissociation. Analyses of the binding data sets conducted using Blitz analysis software (Forte Bio) revealed an equilibrium dissociation constant (K d ) of 94.5 nM, suggesting that both proteins interacted with each other specifically and at a medium intensity (Fig. 2a). TGF-β1 is synthesized as a precursor molecule comprising a signal peptide, LAP, and a mature TGF-β1 polypeptide 18 . Using the same methods, we tested if rh-fortilin interacted with recombinant human latency-associated-peptide (LAP)-TGF-β1 molecule. We found that the binding of fortilin to the LAP-TGF-β1 protein was weaker than that of fortilin to TGF-β1, at 0.93 µM (Fig. S1d).
To gain greater molecular insights into the fortilin-TGF-β1 interaction, we used an SPR-based co-binding/inhibition assay that enables coarse epitope mapping (Fig. 2c) 19 . In this assay, a constant concentration of analyte (TGF-β1) was pre-incubated with increasing concentrations of the ligand-binding domain of the TGF-β1 receptor TGFβRII fused to IgG1 Fc (TGFβRII-Fc), a known interacting partner of TGF-β1. These TGFβ1-TGFβRII-Fc mixtures were injected onto a chip that had been cross-linked with fortilin. If TGFβRII-Fc and fortilin interact with the same epitope on TGF-β1, then we expected that higher TGFβRII-Fc concentrations would reduce the TGF-β1-dependent SPR signal. However, if TGFβRII-Fc and fortilin interact with different TGF-β1 surfaces, then we expected that higher TGFβRII-Fc concentrations would lead to an increased TGF-β1 SPR signal. Importantly, we found that TGFβRII-Fc reduced the TGF-β1 signal in a concentration-dependent manner, indicating that TGFβRII-Fc prevented binding of TGF-β1 to fortilin (Fig. 2c). IgG Fc alone (Fc) did not prevent TGF-β1 from binding fortilin (Fig. S2b, TGFβ1 + Fc), and Fc did not bind fortilin (Fig. S2b, Fc alone). These data are consistent with a model in which fortilin and TGFβRII contact the same surface on TGF-β1, thus providing direct evidence for a mechanism of TGF-β1 inhibition by fortilin.
Fortilin prevents TGF-β1 from activating the Smad-binding element (SBE) in the cell. The above data (Figs. 1, 2), when taken together, suggested that fortilin prevents TGF-β1 from ligating and activating TGFβRII, but we still did not know if this physical interaction has biological and functional significance. To test whether fortilin prevents TGF-β1 from activating the TGF-β1-Smad pathway in the cell, we first expressed strep-tagged fortilin (strep-tag-fortilin) and luciferase (strep-tag-luciferase) using 293T cells, purified them in the sterile condition, characterized them, and found that they were of appropriate sizes and acceptable purity without degradation ( Fig. S3a; Figs. S9, S10). We then transiently transfected HEK293 cells with both the SBE-luciferase plasmid and the Renilla luciferase plasmid (the latter to evaluate transfection efficiency), treated them with TGF-β1 (1 nM) in the presence and absence of strep-tag-fortilin (3 nM), and subjected the cells to the dual luciferase assay as described in the Methods. Fortilin significantly decreased the TGF-β1-induced activation of SBE (luciferase activities; fortilin (−) vs. fortilin (+) = lanes 2 vs. 4 = 8.1 ± 0.2 vs. 5.4 ± 0.5 arbitrary units (A.U.), P < 0.001, N = 4 each) (Fig. 3a).
Next, we used the MFB-F11 cell line to validate these findings using a different cell system. The cell line was originally generated by stably transfecting mouse embryonic fibroblasts from Tgfb1 −/− mice with a construct in which the 12 CAGA boxes-Smad3/ Smad4 binding sequences present in the human plasminogen activator inhibitor-1 gene 20 -are fused to a secreted alkaline phosphatase (SEAP) reporter gene 21,22 . This cell line has been shown to respond to TGF-β1 in a linear fashion from 1 pg/mL to 10 ng/mL (0.04-400 pM), which is an extremely wide dynamic range (ref. 21 and Fig. S3b). We first stimulated the MFB-F11 cells with 156 pM (2 ng/mL) of recombinant TGF-β1 and found that the SEAP activity increased 11.4-fold (Fig. 3b, lanes 1 vs. 2). Both BLI and SPR show the specific interaction between fortilin and TGF-β1. BLI, bio-layer interferometry; K d , dissociation constant; SPR, surface plasmon resonance; His 6 -fortilin, hexa-his-tagged recombinant fortilin; TGFβRIIFc, the ligand-binding segment of the TGF-β receptor II conjugated to the Fc portion of the IgG. a BLI showed the specific binding of fortilin to TGF-β1 at K d of 94.5 nM. b SPR showed that fortilin binds TGF-β1 at K d of 210.5 nM. c After fortilin was conjugated to the SPR chip surface, TGFβRIIFc mixed with various concentrations of TGF-β1 was injected onto the chip surface to evaluate the binding of fortilin to TGF-β1 in the presence of TGFβRII. There was dose-dependent disruption by TGFβRII of the fortilin-TGF-β1 interaction. Because fortilin did not bind TGFβRIIFc (Fig. S2b), the data suggest that fortilin and TGFβRII competitively bind TGF-β1.

Discussion
The most substantial findings of this study are the specific physical interaction between fortilin and TGF-β1 and elucidation of the biological significance of this interaction, namely that fortilin negatively regulates the ability of TGF-β1 to activate its canonical pathway through TGFβRII.
Because the serum concentration of TGF-β1 is drastically higher than those of TGF-β2 and TGF-β3 (Fig. S1a), we focused on the interaction between fortilin and TGF-β1 in the current work. Although the fortilin-TGF-β1 interaction may represent the majority of fortilin-TGF-β interactions in the blood, further investigation is called for to characterize and elucidate the biological significance of the fortilin-TGF-β2/3 interaction.
Although fortilin specifically binds TGF-β1, the fortilin-TGF-β1 interaction is weaker than the interaction between TGF-β1 and TGFβRII. As we previously reported, the K d between TGF-β1 and TGFβRII is 1.49 nM 35 , whereas the K d of the TGF-β1-fortilin interaction is 94.5-210.5 nM (Fig. 2a, b). This suggests that once TGF-β1 binds TGFβRII, fortilin is not capable of disrupting the ligation, which is also consistent with the dose-dependent disruption by TGFβRII of the fortilin-TGF-β1 interaction shown in Fig. 2c. Fortilin may interfere with the biological activities of TGF-β1 by binding and trapping it before it reaches TGFβRII, its receptor.
It has been established that TGF-β binds directly to TGFβRII, which is a constitutional kinase. Bound TGF-β is then recognized by TGFβRI, which is recruited to the TGF-β-TGFβRII complex and becomes phosphorylated and activated by TGFβRII. Activated TGFβRI in turn phosphorylates and activates Smad3. In other words, Smad3 is phosphorylated only when TGFβRI is recruited to the TGF-β-TGFβRII complex and activated by phosphorylation [36][37][38][39] . Our data showing that fortilin inhibits TGF-β1-induced Smad3 phosphorylation (Fig. 4) suggest that fortilin inhibits the recruitment of TGFβRI to TGFβRII by preventing TGF-β1 from ligating TGFβRII but not by preventing the formed TGF-β1-TGFβRII complex from recruiting TGFβR1, as the TGF-β1-TGFβRII interaction is stronger than the TGF-β1fortilin interaction (1.49 nM 35 versus 94.5-210.5 nM, Fig. 2a, b).
Fortilin has no sequence similarity to any of the above TGF-β1interacting proteins, and it possesses several unique attributes that they do not have. First, fortilin is present not only in the extracellular space where it interacts with TGF-β1 but also in the nucleus, ER, and cytosol where it interacts with p53 12 , IRE1α 5 , and MCL1 15 , respectively. Second, fortilin is a multi-functional protein: in addition to binding and inhibiting TGF-β1, fortilin protects cells against apoptosis by (i) binding and inhibiting the pro-apoptotic and tumor suppressor protein p53 12 , (ii) binding and protecting against degradation of the anti-apoptotic Bcl-2 family member protein MCL1 63 , (iii) binding calcium (Ca 2+ ) and preventing Ca 2+ from activating Ca 2+ -dependent apoptosis pathway 64 , and (iv) binding and preventing IRE1α from activating the ER stress-induced apoptosis pathway 5 . Fortilin also binds and potentiates the antioxidant activity of peroxiredoxin-1 7 . Further, fortilin is required for normal cell cycle progression through the stabilization of microtubules 6 . Finally, we do not know if fortilin in the extracellular space has functions other than Fig. 6 Fortilin occupies the binding space for TGF1-β1 that is normally occupied by TGFβRII. +, amino acid residues interfacing between TGF-β1 and TGFBRII; *, amino acid residues interfacing between TGF-β3 and TGFBRII; Identities, the amino acid residues of TGF-β2 and TGF-β3 that are identical to those of TGF-β1; ECS, extracellular space; PM, plasma membrane; P, phosphorylated amino acid residue; TGFβRI, TGF-β1 receptor I; TGFβRII, TGF-β1 receptor II. a Sequence alignment of TGF-β1, -β2, and -β3. Five amino acids (Arg 25 , His 34 , Tyr 91 , Gly 93 , and Arg 94 ) of TGF-β1 (+) and 10 amino acids (Arg 25 , Lys 31 , Trp 32 , His 34 , Lys 37 , Tyr 90 , Tyr 91 , Gly 93 , Arg 94 , and Thr 95 ) of TGF-β3 (*) interface with TGFβRII according to the prior co-crystallization studies. b Computational modeling of dimerized TGF-β1 and either fortilin or TGFβRII. Fortilin (green) and TGFβRII (blue) occupy the same spatial location in relation to dimerized TGF-β1 (red, chains A and B) and either fortilin or TGFβRII, but not both, can bind dimerized TGF-β1 at a given time. binding to and blocking the signal transduction of TGF-β1. We also do not know if TGF-β1 modulates the function of extracellular fortilin. Further investigation is needed to address these questions.
TGF-β1 is involved in complex and diverse biological activities in all organisms. TGF-β1 can suppress early-stage tumors through its potent antiproliferative activity and the induction of cell differentiation and apoptosis 65,66 . In advanced cancer, however, TGF-β1 actually promotes tumor invasion and metastasis by inducing the epithelial-mesenchymal transition, facilitating tumor angiogenesis, and hampering tumor immune surveillance by the host [65][66][67] . TGF-β1 is also implicated in many fibrotic human diseases, including pulmonary fibrosis 68 . Finally, TGF-β1 plays critical roles in the negative regulation of the inflammatory immune response 69 . The results of this study showing the physically and biologically meaningful interaction between fortilin and TGF-β1 represent the beginning of our long-term investigation of the regulatory role of fortilin in numerous TGF-β1mediated biological processes and human diseases, including cancer invasion, metastasis, immune surveillance, and angiogenesis; pulmonary fibrosis and other fibrotic diseases; and inflammatory diseases. Although its activities appear to be expansive and complex, it is also possible that fortilin orchestrates seemingly diverse biological events to achieve a simple cellular goal in response to certain microenvironmental changes.  Images were electronically captured using the Bio-Rad ChemiDoc MP Imaging System, and the signal intensities of protein bands were quantified using Image Lab Software (Bio-Rad, Hercules, CA, USA). To quantify P-Smad3 expression in relation to total proteins loaded, 0.5% (v/v) 2,2,2-trichloroethanol (TCE) was added to a polyacrylamide gel before polymerization. After standard SDS-PAGE, the gel was UV-irradiated on the Bio-Rad ChemiDoc MP Imaging System for 2 min, its image was electronically captured, and the cumulative band densities were calculated to assess loading conditions as previously described 71 . The signal intensity of western blot bands of P-Smad3 was then divided by that of the TCE bands to derive the pSmad3 expression index, which was expressed as A.U. For the quantification of P-Smad3 expression in relation to total Smad3, the signal intensity of P-Smad3 bands quantified by Image Lab Software (Bio-Rad) was divided by that of corresponding total Smad3 bands, which was expressed as A.U.
Generation of recombinant human fortilin. Recombinant human fortilin (rhfortilin) was purified using the Strep-tag purification system (IBA LifeSciences) as described previously 5 with slight modification. Briefly, 293T cells stably expressing human fortilin with an N-terminal Strep-tag II tag (peptide sequence = WSHPQFEK) were collected, washed in PBS, and resuspended in Buffer W (100 mM Tris HCl, pH8, 150 mM NaCl, 1 mM EDTA, and protease inhibitor cocktail, 1 tablet/mL). The cell suspension then was lysed by repeated freeze-thaw cycles and sonicated to shear the genomic DNA. After centrifugation to remove the cell pellet, the cleared cell lysate was passed through a gravity flow Strep-Tactin® XT Superflow® high-capacity column (IBA LIfeSciences). Next, the column was washed five times with Buffer W and eluted with Buffer BXT (Buffer W containing 50 mM biotin). The Strep-tagged fortilin eluent fractions were pooled and concentrated using ultra centrifugal filter units (Amicon™, Millipore Sigma, Burlington, MA, USA). The concentrated proteins were dialyzed in PBS using a Slide-A-Lyzer™ MINI dialysis device (Thermo Fisher Scientific, Waltham, MA, USA). Finally, purified rh-fortilin was characterized by Stain-free total protein TCE® (Bio-Rad) and western blot analyses.
Generation of TGF-β1-rich sera. De-identified human platelet units were obtained from the University of Washington Blood Bank (Seattle, WA, USA) and stored at -80°C until used in the experiment. After the unit was thawed, we added thrombin at the final concentration of 0.5 unit/mL, immediately mixed the solution, incubated it at room temperature for 10 min with gentle shaking, centrifuged the mixture at 10,000 × g. We collected the separated supernatant, labeled it "sera", and stored it at −80°C.
Immunodepletion of fortilin or TGF-β1 from human sera. Mouse anti-TGF-β monoclonal antibody (α-TGF-β mAb, Catalog #: MAB1835-100, Clone 1D11, R&D Systems) and α-fortilin mAb were coupled to Dynabeads® M-280 tosylactivated magnetic beads (Thermo Fisher Scientific) according to the manufacturer's instructions. We incubated 180 µL of the platelet-rich sera with 100 µL each of magnetic beads (20 mg/mL) coupled with either α-TGF-β or α-fortilin mAbs for 2 h at 37°C, immobilized the beads with a magnet, and transferred the supernatants (the sera immunodepleted of fortilin and TGF-β) to fresh microfuge tubes. We repeated the process twice before we stored the samples at -80°C until the ELISA experiments.
ELISA. TGF-β1 ELISA: For ELISA of TGF-β1, the TGF-β1 human ELISA kit was purchased from Abcam (Catalog #: ab100647) and used according to the manufacturer's instructions. We first applied 100 µL of control and study sera to the wells of a 96-well strip plate that had already been pre-coated with α-TGF-β1 capturing antibody. After incubating the plate for 2.5 h at room temperature with gentle shaking and washing it extensively with wash buffer provided by the manufacturer, we added 100 µL of biotinylated α-TGF-β1 detection antibody and incubated it for 1 h at room temperature. After another extensive wash, we added horse-radishperoxidase (HRP)-conjugated streptavidin solution to each well and incubated it for 45 min at room temperature with gentle shaking. After a final extensive wash, we added 3,3′,5,5′-tetramethylbenzidine (TMB) substrate solution to each well, incubated the plate in the dark for 30 min at room temperature with gentle shaking, added 50 µL of stop solution to each well, and obtained the absorbance at 450 nm (Abs 450 ), using the Multilabel Plate Reader (Victor 3 V, Model 1420, Perkin Elmer, Waltham, MA, USA). Fortilin ELISA: For ELISA of fortilin, we first biotinylated αfortilin mAb using the Lighting-Link® Biotinylation Kit (Catalog #: ab201795, Abcam). We then coated the wells of a 96-well plate (Clear Flat Bottom Polystyrene High Bind Microplate, Corning; Catalog #: 9018) by adding 200 µL of 1 µg/mL αfortilin mAb to the wells. We incubated the plate for 1 h at 37°C and then washed it three times with wash buffer (PBS with 150 mM NaCl, 0.05% Tween 20). We then applied 100 µL of control and study sera to the wells of the plate. After incubating the plate for 2.5 h at room temperature with gentle shaking and washing it extensively, we added 100 µL of biotinylated α-fortilin detection antibody and incubated it for 1 h at room temperature. After extensive washing, we added HRPstreptavidin solution (Abcam, Catalog #: ab7403) to each well and incubated the plate for 45 min at room temperature with gentle shaking. Finally, after extensive washing, we added TMB substrate solution (Abcam, Catalog #: ab171522) to each well, incubated the plate in the dark for 3 min at room temperature with gentle shaking, added 100 µL of stop solution (Abcam, Catalog #: ab171529) to each well, and obtained the absorbance at 450 nm. Modified ELISA to detect the fortilin-TGF-β1 interaction in vivo: To evaluate the in vivo complexing between TGF-β1 and fortilin, we first applied 100 µL of these samples to the wells of a 96-well TGF-β1 human ELISA strip plate (Abcam, Catalog #; 100647) that had already been precoated with α-TGF-β1 capturing antibody. After incubating the plate for 2.5 h at room temperature with gentle shaking and washing it extensively, we added 100 µL of biotinylated α-fortilin detection antibody and incubated it for 1 h at room temperature. After extensive washing, we added HRP-streptavidin solution to each well and incubated the plate for 45 min at room temperature with gentle shaking. Finally, after extensive washing, we added TMB substrate solution (Abcam) to each well, incubated it in the dark for 30 min at room temperature with gentle shaking, added 50 µL of stop solution (Abcam) to each well, and obtained the absorbance at 450 nm.
BLI analysis. The interaction between fortilin and TGF-β1 was evaluated using the BLItz System (ForteBio) as described previously 5 . First, rh-fortilin protein was biotinylated by mixing it with 2 mM NHS-PEG4-biotin solution and incubating the mixture for 30 min at room temperature. The biotinylated protein was then purified to remove free biotin using Zeba™ Spin Desalting Columns (Thermo Fisher Scientific). Next, the biotinylated fortilin was immobilized on streptavidin-coated biosensors (ForteBio) at a concentration of 30 μg/mL in PBS for 600 s. After fortilin-coated biosensors were buffer-exchanged in PBS for 30 s, various concentrations of rh-TGF-β1 (OriGene; Catalog #: TP300973) were added for 180 s to evaluate the association between the two proteins. Finally, rh-TGF-β1 protein solution was replaced by PBS for 180 s to evaluate their dissociation. The binding data were processed, and a dissociation constant (K d ) was calculated using BLItz analysis software based on two independent experiments. The same procedure was performed for the LAP-TGF-β1 protein (R&D Systems), except the K d was calculated based on three independent experiments. SPR analysis. All experiments were performed using a Biacore 3000 (Uppsala, Sweden) and carried out at 25°C using HBS-EPS (0.01 M HEPES, 0.5 M NaCl, 3 mM EDTA, 0.005% (v/v) Tween 20, pH 7.4) as running buffer. The experimental flow rate was 20 μL/min. Different forms of rh-fortilin were immobilized onto three flow channels (FCs) of a CM5 chip using amine-coupling chemistry as follows: FC2: 7060 RU N-His 6 -fortilin; FC3: 3223 RU C-His 6 -fortilin; FC4: 3667 RU streptag-fortilin. To scout for binding conditions, rh-TGF-β1 produced in-house was injected at a concentration of 100 nM over each channel. To obtain binding rates, rh-TGF-β1 was injected over N-His 6 -fortilin at the indicated concentrations. Binding rates were calculated by fitting data to a 1:1 Langmuir interaction model with mass transport limitation using BiaEvaluation software (Biacore). K d s were determined by calculating the ratio of binding and dissociation rate constants. For inhibition analysis, 100 nM rh-TGF-β1 was pre-incubated with indicated concentrations of the inhibitor rh-TGFβRII-Fc and injected over experimental and control flow channels 19 . After each binding cycle, the fortilin-coupled surface was regenerated to base line by injecting 20 µL of 1 M NaCl.
Luciferase assay using HEK293 SBE-Luc cells. We transiently transfected HEK293 cells with the SBE-luciferase vector (pGL4.48 [luc2P/SBE/Hygro], Promega, Madison, WI, USA) along with the Renilla luciferase control reporter vector (pRL, Promega), which allowed us to normalize firefly luciferase activities according to transfection efficiency. The following day, we treated the transfected HEK293 cells with rh-TGF-β1 (1 nM) or vehicle (PBS) in the presence and absence of step-tagfortilin (3 nM) in quadruplicate, incubated the samples for 18 h at 37°C, and subjected the cells to the Dual-Glo-Luciferase Assay System (Promega) according to the manufacturer's instructions. We calculated relative luciferase activity (FLU/ RLU) by dividing firefly luciferase units (FLU) by Renilla luciferase units (RLU) and expressed the results as A.U. We then normalized FLU/RLU of the cells treated with rh-TGF-β1 to that of the cells treated with vehicle in either the presence or absence of fortilin.
SEAP reporter assay using MFB-F11 SBE-SEAP cells. MFB-F11 cells 21,22 were a kind gift from Dr. Tony Wyss-Coray (Stanford University, Stanford, CA, USA). The cells were generated at the Wyss-Coray lab by stably transfecting mouse embryonic fibroblasts from Tgfb1 −/− mice with a synthetic promoter element containing 12 CAGA boxes fused to a SEAP reporter gene 22 . The cells were kept and propagated in DMEM supplemented with 10% FBS. For SEAP reporter assays, we first acclimated murine MFB-F11 cells in FibroLife Fibroblast Serum Free Medium (LifeLine Cell Technology, Oceanside, CA, USA) with 1% FBS. We seeded MFB-F11 cells in a 96-well plate at 50,000 cells per well (N = 6 each) and incubated them for 24 h. The next morning, we replaced the medium with 100 µL of FibroLife without FBS and incubated the cells at 37°C for 2 h. We then stimulated the cells with recombinant mouse (rm) TGF-β1 (R&D Systems, Catalog #: 7666-MB/CF; 156 pM or 2000 pg/mL) pre-incubated with strep-tagged rh-fortilin (either 19.5 or 195 nM), strep-tagged recombinant luciferase (either 19.5 or 195 nM), or anti-TGF-β1 monoclonal antibody (R&D Systems, Catalog #: MAB1835-100, Clone: 1D11) for 24 h at 37°C. We then subjected 40 µL of supernatants to the Phospha-Light™ SEAP Reporter Gene assay (Thermo Fisher Scientific, Catalog #: T1015) according to the manufacturer's instructions.
Smad3 phosphorylation assay by western blot analysis and ELISA. We seeded 0.7 × 10 6 FibroLife-acclimated MFB-F11 SBE-SEAP cells in each well of 6-well plates and incubated them for 24 h. We then replaced the medium with 100 µL of FibroLife without FBS and incubated the cells at 37°C for 2 h. Next, we stimulated the cells for 45 min at 37°C with rm-TGF-β1 (2 ng/mL or 156 pM) that had been pre-incubated with either fortilin or control luciferase proteins. We harvested the cells for either standard western blot analysis or the P-Smad3 (pS423/S425) ELISA, with the latter following the manufacturer's instructions (Abcam, Catalog #: ab186038).
Computational molecular docking of fortilin, TGF-β1, and TGFβRII. Threedimensional protein structural models of fortilin (PDB ID: 2HR9), TGF-β1 (PDB ID: 3KFD), and the TGF-β1:TGFβRII complex (PDB ID: 5TY4) for docking experiments were obtained from the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (PDB) 29,[74][75][76] . Protein-protein docking of fortilin with TGF-β1 and TGF-β1:TGFβRII was performed using the ClusPro server utilizing the PIPER docking algorithm 33,34,77,78 . For the docking procedures, the center-of-mass (COM) of the receptor TGF-β1 protein (or TGF-β1:TGFβRII) was fixed, and the fortilin position was sampled at 70,000 possible rotational orientations about the COM. At each rotational position, the ligand translational position was sampled at a resolution of 1 Å to find the corresponding lowest-energy conformation. In total, the algorithm sampled~10 9 possible relative orientations of the proteins. Interaction energies (or scores) for each conformation were calculated using an electrostatic-favored weighting due to the overall high density of charged residues on the protein surfaces. The docking procedure is followed by clustering of similarly oriented structures of the 1000 lowest-energy structures. The interfacial root mean squared deviation (IRMSD) of backbone atoms are calculated for each structure, and that with the largest number of neighboring structures within a 9 Å radius is defined as the center of the first cluster. All protein-protein orientations within the 9 Å IRMSD are considered part of the first cluster and removed from the population. The procedure is repeated for the remaining population until all docked structures are clustered. The populations of the resulting clusters are proportional to the thermodynamic probabilities of finding the proteins in each specific binding orientation. The scores are calculated from molecular mechanics (CHARMM force field) for the structures obtained from docking.
Statistics and reproducibility. The degree of the spread of data was expressed by the standard deviation (mean ± SD). Student's t-test was used to compare the means of two groups. To compare the means of three groups, we used one-way ANOVA with Fisher's pairwise comparison. P < 0.05 was considered to be statistically significant. P < 0.10 was considered to show a trend toward statistical significance. The numbers of biological replicates used in in vivo experiments were determined by (i) power analysis, assuming an α error rate of 0.05, β error rate of 0.20, and expected difference of 25% and using Minitab 17 (State College, PA, USA) or (ii) our previous dataset and experience from similar experiments performed in the past.