Transforming growth factor β1 (TGFβ1)-induced CD44V6-NOX4 signaling in pathogenesis of idiopathic pulmonary fibrosis

Idiopathic pulmonary fibrosis (IPF) is a progressive clinical syndrome of fatal outcome. The lack of information about the signaling pathways that sustain fibrosis and the myofibroblast phenotype has prevented the development of targeted therapies for IPF. Our previous study showed that isolated fibrogenic lung fibroblasts have high endogenous levels of the hyaluronan receptor, CD44V6 (CD44 variant containing exon 6), which enhances the TGFβ1 autocrine signaling and induces fibroblasts to transdifferentiate into myofibroblasts. NADPH oxidase 4 (NOX4) enzyme, which catalyzes the reduction of O2 to hydrogen peroxide (H2O2), has been implicated in the cardiac and lung myofibroblast phenotype. However, whether CD44V6 regulates NOX4 to mediate tissue repair and fibrogenesis is not well-defined. The present study assessed the mechanism of how TGF-β-1-induced CD44V6 regulates the NOX4/reactive oxygen species (ROS) signaling that mediates the myofibroblast differentiation. Specifically, we found that NOX4/ROS regulates hyaluronan synthesis and the transcription of CD44V6 via an effect upon AP-1 activity. Further, CD44V6 is part of a positive-feedback loop with TGFβ1/TGFβRI signaling that acts to increase NOX4/ROS production, which is required for myofibroblast differentiation, myofibroblast differentiation, myofibroblast extracellular matrix production, myofibroblast invasion, and myofibroblast contractility. Both NOX4 and CD44v6 are up-regulated in the lungs of mice subjected to experimental lung injury and in cases of human IPF. Genetic (CD44v6 shRNA) or a small molecule inhibitor (CD44v6 peptide) targeting of CD44v6 abrogates fibrogenesis in murine models of lung injury. These studies support a function for CD44V6 in lung fibrosis and offer proof of concept for therapeutic targeting of CD44V6 in lung fibrosis disorders.


Idiopathic pulmonary fibrosis (IPF) is a progressive clinical
syndrome of fatal outcome. The lack of information about the signaling pathways that sustain fibrosis and the myofibroblast phenotype has prevented the development of targeted therapies for IPF. Our previous study showed that isolated fibrogenic lung fibroblasts have high endogenous levels of the hyaluronan receptor, CD44V6 (CD44 variant containing exon 6), which enhances the TGF␤1 autocrine signaling and induces fibroblasts to transdifferentiate into myofibroblasts. NADPH oxidase 4 (NOX4) enzyme, which catalyzes the reduction of O 2 to hydrogen peroxide (H 2 O 2 ), has been implicated in the cardiac and lung myofibroblast phenotype. However, whether CD44V6 regulates NOX4 to mediate tissue repair and fibrogenesis is not well-defined. The present study assessed the mechanism of how TGF-␤-1-induced CD44V6 regulates the NOX4/reactive oxygen species (ROS) signaling that mediates the myofibroblast differentiation. Specifically, we found that NOX4/ROS regulates hyaluronan synthesis and the transcription of CD44V6 via an effect upon AP-1 activity. Further, CD44V6 is part of a positivefeedback loop with TGF␤1/TGF␤RI signaling that acts to increase NOX4/ROS production, which is required for myofi-broblast differentiation, myofibroblast differentiation, myofibroblast extracellular matrix production, myofibroblast invasion, and myofibroblast contractility. Both NOX4 and CD44v6 are up-regulated in the lungs of mice subjected to experimental lung injury and in cases of human IPF. Genetic (CD44v6 shRNA) or a small molecule inhibitor (CD44v6 peptide) targeting of CD44v6 abrogates fibrogenesis in murine models of lung injury. These studies support a function for CD44V6 in lung fibrosis and offer proof of concept for therapeutic targeting of CD44V6 in lung fibrosis disorders.
Idiopathic pulmonary fibrosis (IPF), 3 the hallmark of impaired repair of lung injury, is characterized by persistent activation and differentiation of resident fibroblasts to myofibroblasts (2,3). However, the underlying etiology in IPF and the progressive nature of the fibrotic process in IPF are generally unknown (3).
Accumulated evidence indicates that the increased oxidative stress associated with an oxidant-antioxidant imbalance and elevated reactive oxygen species (ROS) production have a key role in the lungs of patients with IPF (4 -6). NADPH oxidases (NOXs) are a family of flavoenzymes that are unique in that they are solely responsible for primary ROS (H 2 O 2 ) production (7,8). These multisubunit protein complexes consist of five subunits: a membrane-associated cytochrome b 558 containing gp91 phox and p22 phox and a cytosolic complex of p40 phox , p47 phox , and p67 phox . The family of human NOX proteins consists of five members (NOX1-to -5) and two related dual oxidases (DUOX1 and -2). NOX1, NOX2, and NOX3 show a high degree of homology and require p40 phox , p47 phox , and p67 phox subunits and activation by Rac protein. In addition, the p22 phox subunit is important for the stability and functioning of NOX1 to NOX4 (9 -11). NOX5 (absent in mice) (12) is not regulated by p22 phox (11,13). Recent studies have reported that specific NOX (NOX2 and NOX4) and DUOX isoforms (DUOX1 and DUOX2), and subunits p22 phox and p47 phox are involved in the pathogenesis of obstructive lung disorders and that they correlate with an enhanced H 2 O 2 production (14). NOX enzymes are critical determinants of the redox status of the lung tissue and have a critical role in the myofibroblast activation in IPF disease (6,15,16).
Fibrosis is caused by unrestrained wound-healing responses during tissue repair in pathological conditions, including IPF (17). During wound repair, both fibroblasts and myofibroblasts invade the wound and synthesize a variety of extracellular matrix (ECM) proteins and the glycosaminoglycan hyaluronan (HA) in the fibrotic tissue (18,19). At sites of tissue damage and wound healing, fibroblasts with a contractile phenotype are essential for the synthesis of the collagen-rich scar and for providing the force for wound contraction (20). Fibrotic scars are characterized by accumulation of contractile matrix proteins produced by differentiation of fibroblasts to myofibroblasts, which can then compromise normal tissue functions. One of the major profibrotic cytokines involved in fibrogenesis is TGF␤1, as shown in the pathogenesis of IPF (21,22). In addition to the direct effect of TGF␤1 on ECM turnover, it can drive fibroblast-to-myofibroblast differentiation and activation, which is associated with disease pathologies (2,(23)(24)(25)(26)(27). In fibrogenic circumstances, the myofibroblasts comprise the crucial cell population (2). In contrast to wound healing, however, a persistent accumulation of myofibroblasts is associated with the pathological reorganization and expansion of ECM components that may not normally be present or are present at only low levels in normal tissue (28).
One of the earliest fibrogenic responses is the deposition of HA in the pericellular matrix surrounding the differentiated cells (29 -32). HA is synthesized by HA synthase enzymes, of which three distinct genes have been isolated and characterized, HAS1, HAS2, and HAS3 (33). HA turnover occurs under normal physiologic settings, and the increase in HA levels during pathologic conditions reveals the balance between its synthesis and catabolism. Hyaluronidases are enzymes that degrade HA, and two distinct hyaluronidase genes (HYAL1 and HYAL2) have been reported to be involved in the intracellular and extracellular catabolism of HA in a CD44-dependent manner (34). It has been shown that lung tissue from human IPF patients and from mice in which lung fibrosis was induced by bleomycin share a potentially important clue as to the cause of fibrosis, the overexpression of Has2 (hyaluronan synthase 2) by myofibroblasts, the enzyme that can synthesize HA (19). A recent study also provides evidence that deletion of Has2 in mouse mesenchymal cells increased the cellular senescence of fibroblasts in bleomycin-induced mouse lung injury (35). The authors also suggest that targeting Has2 to induce fibroblast senescence could be an attractive approach to resolve tissue fibrosis (35). On the other hand, studies also show that senescent human fibroblasts also resist apoptosis caused by growth factor deprivation and oxidative stress, indicating that resistance to apoptosis might partly explain why senescent and apoptosis-resistant myofibroblasts are highly stable in culture and induce persistent fibrosis in lungs of aged mice (6,36,37). Myofibroblast apoptosis is crucial to the regression of fibrotic scars and the regeneration of healthy tissue during wound repair and may be aberrant in diseases such as IPF. It is now recognized that myofibroblasts have several origins, contribute significantly to connective tissue remodeling by exerting contractile forces and producing ECM components, regress and disappear by apoptosis on wound epithelialization, and may persist in fibrotic situations and cause organ dysfunction (38).
CD44, a major receptor for HA, is involved in intracellular signaling (1, 39 -56) and is an avid regulator of apoptosis in fibroblasts (1,57). Increased synthesis of HA and expression of CD44 have been detected in numerous fibrotic conditions associated with organ dysfunction (58 -61). CD44 is known to have a major role in regulating cell-cell adhesion (49,62), migration (44,49,(63)(64)(65)(66), differentiation (1,67,68), and proliferation (69), and it therefore has an important role in wound healing (50). The human CD44 gene contains 19 exons (70). Exons 6 -14 are alternatively spliced to produce CD44 variant isoforms (70). CD44 is important in maintaining the integrity of the fibroblast actin cytoskeleton (64,(72)(73) and in facilitating an organized, directional migratory response to injury (74). CD44V6, in which exon 6 is expressed, is of particular interest because interaction of HA with CD44V6 alters cellular function in response to various growth factors and cytokines (75). Our recent study showed that a unique functional activity of CD44V6 includes stimulation of ERK activation, increased collagen-1 (COL1A1) synthesis, and induction of ␣-smooth-muscle actin (␣-SMA) in isolated lung myofibroblasts through TGF␤1 autocrine signaling in fibrogenic lung disease of human subjects (1). Studies using IPF lung-derived fibroblasts (IPFFbs) demonstrated that TGF␤1-dependent profibrotic responses, including resistance to apoptosis, up-regulation of ␣-SMA gene expression, and the secretion of the ECM components fibronectin and COL1A1, are NOX4-dependent and contribute to persistent fibrosis in IPF disease (6,15,16). A recent study also indicates that the expression of CD44 is regulated in a NOX-dependent manner in atherosclerosis disease (76). In addition, ROS can degrade HA under oxidative conditions (77). Thus, understanding myofibroblast activation is crucial to understanding this aspect of injury and repair. Thus, we hypothesize that CD44V6 may be a critical determinant of the fate of fibrosis by promoting myofibroblast activation through its interaction with NOX4 in the fibroblasts from IPF patients and that this leads to the progressive fibrosis.
Proinflammatory cytokines, such as tumor necrosis factor ␣ (TNF␣), are key features of inflammatory processes and may CD44V6, NOX4, and TGF␤1 signaling in IPF cause tissue damage and destruction (78). TNF␣ acts in vasodilation, edema, leukocyte adhesion to the epithelium, macrophage activation, fever, and metalloprotease activation and contributes to the oxidative stress at inflammation sites (79,80). Recent observations also indicate that eosinophils found in fibrotic lung express elevated levels of cytokines known to be important in lung fibrosis. These findings suggest a possible role for TNF␣ in eosinophil recruitment and cytokine expression in this disease. Eosinophil-derived cytokines include MCP-1 (monocyte chemotactic factor-1) and TGF␤1. This cytokine networking orchestrated by TNF␣ could, in turn, amplify the inflammatory response and drive the progression to fibrosis and end-stage lung disease (81). Furthermore, TNF␣ inhibits the anti-inflammatory cytokine interleukin-10 (IL-10) production (82), triggering chronic inflammation (83) and autoimmune diseases (84). Recent studies suggest that cells of the monocyte/macrophage lineage are also present, especially at the early stages of the fibrosis disease (85). Infiltrating leukocytes are potential sources of a number of immunomodulatory or profibrotic cytokines. It seems likely that complex cytokinemediated networks of interaction between different cell types are involved in the development of tissue fibrosis. As a consequence of cytokine stimulation, an activated population of fibrogenic fibroblasts deposit excessive extracellular matrix, which is the pathological hallmark of the established disease (86).
The pleiotropic cytokine TGF␤1 and its signaling activation of downstream molecules has been proposed as a control mechanism for the induction of inflammatory cells and molecules. TGF␤1 binds directly to TGF␤ receptor II (TGF␤RII), which is a constitutively active kinase. Bound TGF␤1 is then recognized by receptor I (TGF␤RI), which is recruited into the complex and becomes phosphorylated by TGF␤RII. Phosphorylation allows TGF␤RI to propagate the signal to downstream substrates, including SMAD2 and SMAD3. This provides a mechanism by which a cytokine can generate the first step of a canonical or a non-canonical signaling cascade (87,88). The growing interest in the TGF␤1 pathway as a treatment option for inflammatory diseases, especially due to the possibility of controlling reactive inflammatory cells, led us to the development of novel TGF␤1-like peptides from the CD44V6 exon sites that could outperform the effects of TGF␤1.
Using fibroblasts from normal human lungs, we show that sustained production of CD44v6 involves TGF␤1/SMAD3-mediated NOX4 induction and ROS production and that NOX4 regulates CD44v6 gene expression through transcription factor AP-1, which then regulates Has2 mRNA transcription in murine lung fibroblasts. In addition, our data indicating that NOX4 regulates hyaluronidase expression in isolated mouse normal lung fibroblasts (MNLFbs) treated with TGF␤1 is significant because increased HA during patho-physiologic conditions reflects the balance between its synthesis and its catabolism. Genetic targeting of CD44v6 and an endogenous CD44v6 competing peptide (V6-PEP) tested in leukocytes significantly down-modulated fibrogenic fibroblast-mediated transendothelial migration of leukocytes induced by TNF-␣. Importantly, our CD44v6 shRNA and V6-PEP peptide provide a preferred balanced immune response, with increased IL-10, which is essential for the immune response modulation, especially when associated with decreased TNF-␣ and abrogation of leukocyte-endothelial interactions, which may prevent excessive inflammation and tissue damage (89). Our proof-of-concept data suggest that non-toxic/anti-inflammatory CD44V6 shRNA (V6 shRNA)/transferrin (Tf)-PEG-PEI (nanoparticles) or endogenous V6-PEP/nanoparticles abrogate fibrogenesis in a murine model of lung injury by reducing NOX4/ROS activity, reducing TNF-␣ production, and stimulating IL-10 production, with subsequent abrogation of transendothelial migration of leukocytes.

TGF␤1-induced CD44V6 synthesis regulates NOX4 and NOX4-regulated H 2 O 2 during activation of human and mouse mesenchymal lung fibroblasts
The ability of TGF␤1 to stimulate cellular production of reactive oxygen species (H 2 O 2 ) through NOX in human and mouse lung fibroblasts has been established (16), and NOX is activated through CD44 in an atherosclerosis model (90). Therefore, we investigated the role of NOX in TGF␤1-induced regulation of CD44V6. Fig. 1A shows that expression of all of the components of the NOX pathway was found in cultures of IPFFbs. Fig. 1B shows that NOX4 mRNA is highly expressed in human normal lung fibroblasts (HNLFbs) stimulated with TGF␤1, whereas other members of the NOX gene family were not affected at the mRNA level.
To define the specific role of TGF␤1-induced CD44V6 on the regulation of NOX4, we treated HNLFb cultures with CD44V6 siRNA. Fig. 1C shows that CD44V6 siRNA efficiently blocked NOX4 induction by TGF␤1. Further, NOX4 siRNA also inhibited the induction of NOX4 by TGF␤1 (Fig. 1C). Because the TGF␤1-induced NOX4 synthesis has been largely attributed to SMAD3 signaling (16), we determined whether SMAD3 is required for H 2 O 2 generation in HNLFbs. Fig. 1D shows that NOX4 knockdown greatly inhibits the TGF␤1-induced synthesis of H 2 O 2 , as does the siRNA knockdown of CD44V6 and SMAD3 (Fig. 1E), to levels equivalent to or better than treatment with GKT137831, a small molecule NOX4/ NOX1 inhibitor from Genkyotex, Switzerland (GKT; Fig. 1D).
Our results in Fig. 2 (A and B) validated the use of these three siRNAs for the in vivo experiments used in this study. Fig. 2A shows that the elevated levels of TGF␤1-induced NOX4 mRNA are significantly reduced by both CD44V6 siRNA and NOX4 siRNA, which are increased to near-control levels by the respective cDNA treatments in IPFFbs. Fig. 2B shows that H 2 O 2 release is inhibited by CD44V6 siRNA and SMAD3 siRNA, which are increased to control level by CD44V6 cDNA and partially increased by SMAD3 cDNA. In the experiments in Fig. 2 (C and D), we confirmed that these siRNAs (and their respective shRNAs) used alone or used with Tf-PEG-PEI (nanoparticles) (for better transfection) do not have any significant off-target effects in the in vivo settings by comparing the effects of poly(I:C) with the siRNAs and shRNAs on IFN␣ secretion in C57BL/6 mice (Fig. 2, C and D). The shRNAs were carefully tested for off-target effects in in vitro settings in Fig. 12 (G and H) of our companion paper (71). Synthetic siRNA duplexes or poly(I:C) were administered as i.p. injections using "low pressure" (low-pressure tail vein (LPTV); 1% (v/w)). Plasma collected at 2 h was analyzed by ELISA for IFN␣ secretion levels. The strong induction by poly(I:C) was absent with the treatments with siRNAs against NOX4, CD44V6, or SMAD3 and with the shRNAs against NOX4 or CD44V6. RNase-treated poly(I:C) was used as a control. Fig. 2D shows that nanoparticles for V6 shRNA, NOX4 shRNA, or V6-PEP injected (i.p. or by tail vein) with low pressure (LPTV; 1% (v/w)) or "high pressure" (HPTV; 10% (v/w)) also do not induce plasma IFN␣. Further, we verified the blocking of CD44V6, or SMAD3 by specific shRNAs. The shRNA experiments were corroborated with a knockdown or with a rescue experiment. We verified the blocking of CD44V6 or SMAD3 by specific shRNAs for the coding sequences (CDS), by co-transfecting the shRNA for the target gene with or without corresponding cDNA transfection, or by replacing the knocked down gene(s) (i.e. a gene-replacement strategy, designed to circumvent the shRNA knockdown). This is accomplished by the indicated shRNA-mediated knockdown and corresponding knock-in (KI) gene transfection (see Table 1 and Fig. 12 (A-F) and the accompanying legend in our companion paper (71).
To define the in vivo role of NOX4 in the fibrogenic and reparative response to injury of the murine lung, we used a murine model of bleomycin instillation that causes epithelial injury with consequential TGF␤1 up-regulation, myofibroblast activation, and lung fibrosis (91). We identified Nox4 as one of the most highly induced genes in murine lung myofibroblasts isolated from lung tissues from 21-day bleomycin-treated mice (21dBLMFbs) compared with MNLFbs from lung tissues of PBS-treated mice (Fig. 3A). We then determined whether Nox4 protein expression was induced in fibroblasts isolated from the fibrogenic and reparative phases of the bleomycin-induced lung injury. Nox4 protein expression was induced in a time-de- A, mRNA expressions of NOX components are shown in HNLFbs and IPFFbs. B, mRNA expressions by real-time PCR for NOX1, NOX2, and NOX4 in NLFbs treated with or without TGF␤1 and expressed relative to ␤-actin. C, effects of control siRNA, NOX4 siRNA, and CD44V6 siRNA on NOX4 mRNA expression in HLFbs treated with or without TGF␤1 are shown and expressed relative to ␤-actin. D, effects of inhibiting NOX4 by NOX4-specific siRNA or by pharmacologic inhibitor GKT137831 (1 M) on extracellular release of H 2 O 2 by HLFbs treated with or without 2.5 ng/ml TGF␤1 for 48 h. E, effects of blocking CD44V6 and SMAD3 using specific siRNAs on extracellular release of H 2 O 2 by HNLFbs treated with or without 2.5 ng/ml of TGF␤1 for 48 h. The data in A are representative of three independent experiments. The experimental data in B-E are from three sets of HNLFbs with three independent experiments and are expressed as means Ϯ S.D. (error bars). Statistical analysis was with ANOVA (B, *, p Յ 0.005 versus TGF␤1-untreated control group; C, *, p Յ 0.01 versus TGF␤1-treated control shRNA group; D and E, *, p Յ 0.01 versus TGF␤1-treated control siRNA group). pendent manner, increasing from day 7 up to day 21 and then returning to near-baseline levels at day 45 (Fig. 3B). Similar induction of CD44v6 expression was observed during these fibrogenic phases (see Figs. 2 (A and B) and 3F in our companion paper (71)), supporting cross-talk between Nox4 and CD44v6 expression that can lead to myofibroblast activation and fibrosis after lung injury. However, expression of Nox2, which is predominantly expressed in phagocytic cells, was increased on day 14 and returned to near-baseline levels at day 21 during the fibrogenic phase and at day 45 after bleomycin injury, when inflammatory responses had subsided (Fig. 3B).
The myofibroblast phenotype contributes to the healing of tissue injury by secreting ECM proteins by remodeling and contracting the ECM (16,92,93). Fig. 4A shows that by silencing Nox4 or CD44v6, the ␣-Sma, fibronectin, and pro-collagen-1 proteins, as well as Nox4 and CD44v6 proteins, are decreased in 21dBLMFbs (Fig. 4A) and IPFFbs (Fig. 4B). Interestingly, Nox4 regulates CD44v6 expression and vice versa in 21dBMLFbs and in IPFFbs (Fig. 4, A and B). In addition, we confirmed the knockdown experiments, where we always use shRNAs for CDS, or shRNAs against noncoding sequences (NCDS) (designed to target the 3Ј-untranslated region (UTR) of the mRNA) as ideal negative control. Total cell lysates were examined by Western blot analysis for the indicated proteins expressed relative to ␤-tubulin. The data in the experiments are from three sets of 21dBLMFbs with three independent experiments. The densitometry results obtained for Western blots of the indicated proteins are shown in bottom panels of the experiments in Fig. 4A. We also verified the effect of the Nox4 shRNA and CD44v6 shRNA on the ␣-SMA, fibronectin, and pro-collagen-1 proteins, as well as the NOX4 and CD44v6 proteins, with rescue of the observed shRNA-mediated knockdown phenotype by expression of a resistant form of the targeted mRNA. This is performed by transfecting the cells with specific shRNAs for the CDS of the target gene, by co-transfecting the shRNA for the target gene with or without corresponding cDNA transfection, or by indicated shRNA mediated knockdown and A, IPF fibroblasts expressing control siRNA, NOX4 siRNA, or CD44V6 siRNA were further transfected with NOX4 cDNA or CD44V6 cDNA or treated with or without 2.5 ng/ml TGF␤1 for 12 h to induce NOX4 mRNA expression. Total RNAs were examined by real-time PCR analysis for NOX4 mRNA expressed relative to ␤-actin (A, *, p Յ 0.005 versus control siRNA-transfected group). B, IPFFbs expressing control siRNA, CD44V6 siRNA, or SMAD3 siRNA were further transfected with CD44V6 cDNA or SMAD3 cDNA or treated with or without 2.5 ng/ml TGF␤1 for 48 h to induce extracellular release of H 2 O 2. The averages of three replicate fibroblast cultures are presented, and error bars represent S.D.; *, p Յ 0.005. C, plasma IFN-␣ induction levels in female C57BL mice are shown. 2.5 mg/kg (unless otherwise indicated) of nucleic acid was injected through the tail vein by high pressure (10% (v/w)), and plasma was collected 2 h after injection. The averages of three replicate mice are presented, and error bars represent S.D.; *, p Յ 0.005. D, effects of administration of D5W (5% (w/v) glucose in water), control shRNA/nanoparticle, poly(I:C), CD44V6 shRNA/nanoparticle, NOX4 shRNA/nanoparticle, or V6-PEP/nanoparticle are shown. Mice received 2.5 mg/kg nucleic acid either intraperitoneally (IP) or through the tail vein via LPTV (1% (v/w)) or HPTV (10% (v/w)). The averages of three replicate mice are presented, and error bars represent S.D.; *, p Յ 0.005. E and F, the shRNA experiments were corroborated with a knockdown or with a rescue experiment. We verified the blocking of CD44V6 or SMAD3 by specific shRNAs for the CDS, by co-transfecting the shRNA for the target gene with or without corresponding cDNA transfection, or by replacing the knocked down gene(s) (i.e. a gene-replacement strategy, designed to circumvent the shRNA knockdown); this is accomplished by indicated shRNA-mediated knockdown and corresponding KI gene transfection (see Table 1  Redox-dependent activation of IPF mesenchymal cells by NOX4 is known to be involved in regulation of myofibroblast activation during lung fibrosis (16). To determine the role of CD44V6 in conjunction with NOX4, we examined lung tissue sections from human subjects with IPF. Both NOX4 and CD44V6 are highly expressed in cells that are presumably from inflammatory areas of the injured IPF lung (Fig. 5A). We then examined lung fibroblasts isolated from explants of IPF lung tissue (IPFFbs). Fig. 5B shows that in IPFFb TGF␤1 also induces CD44V6 and NOX4 proteins, which in turn are necessary for induction of ECM proteins fibronectin and pro-collagen-1.
This result also indicates that basically IPFFbs respond the same way to TGF␤1 as do NLFbs. CD44V6 and NOX4 are also required for the induction of ␣-SMA, COL1A1, and fibronectin mRNAs (Fig. 5, C-G) and for their protein expression (Fig. 5B). Similar to our findings in 21dBLMFbs, both CD44V6 and NOX4 regulate each other's mRNA expression (Fig. 5, F and G) and protein expression (Fig. 5B). Fig. 6A shows that shRNA-mediated knockdown of CD44V6 and NOX4 inhibited serum-stimulated proliferation of IPFFbs to the same extent as the ROS inhibitor, N-acetyl-L-cysteine (NAC). Fig. 6B shows that the CD44v6/NOX4/ROS/AKT pathway inhibits MMP2 and MMP9 expression. Consistent with this result, the suppression of NOX4, CD44v6, or ROS significantly decreased cell invasion of IPFFbs compared with the control cells (Fig. 6D). The results in Fig. 6C validated the knockdown effect of NOX4 shRNA and CD44V6 shRNA by comparing these shRNAs (CDS) with those of NCDS transfection in human IPFFbs. In summary, the data in Figs. 1-6 support the key role for the CD44V6-NOX4 pathway in myofibro- . In this experiment, the shRNA experiments were corroborated with a knockdown or with rescue of the observed shRNA-mediated knockdown phenotype by expression of a resistant form of the targeted mRNA. This is performed by transfecting the cells with specific shRNAs for the CDS of the target gene, by co-transfecting the shRNA for the target gene with or without corresponding cDNA transfection, or by replacing the knocked down gene(s) (i.e. a gene-replacement strategy, designed to circumvent the shRNA knockdown). This is accomplished by the indicated shRNA-mediated knockdown and corresponding KI gene transfection (see Fig. 12 (A-G) and the Fig. 12 legend in our companion paper (71)). Total cell lysates were examined by Western blot analysis for the indicated proteins expressed relative to ␤-tubulin. The data in the experiments are from three sets of IPFFbs with three independent experiments. The densitometry results obtained for Western blots of the indicated proteins are shown in the bottom panels of the experiments in B. A and B, densitometry graphs, statistical analysis with ANOVA; *, p Յ 0.005 versus control shRNA (Contsh)-transfected group; NS, statistically nonsignificant. Error bars, S.D.
blast invasion, proliferation, and secretion of ECM and expression of contractile proteins necessary for differentiation of human IPFFbs.

CD44v6 regulates TGF␤RI kinase activity to induce downstream TGF␤1 signaling in IPF lung disease
HA has been reported to induce ROS in a CD44-dependent manner (94,95), and NOX mediates increases in ROS production (16). Therefore, signaling events that mediate CD44v6 and TGF␤1 are important in lung fibrosis of human subjects (1). In this study, we demonstrated the association of TGF␤RI and CD44v6 in the Duolink assay of IPFFbs and its near absence in HNLFbs (Fig. 7A). Therefore, we determined whether the interaction between CD44V6 and the TGF␤1 receptor(s) has a significant role in regulating fibrogenic fibroblast cell-specific behaviors, such as p-SMAD activation and NOX4 production. The immunoblots in Fig. 7B from HNLFb cultures show that overexpression of HAS2 significantly up-regulates phosphoser-

CD44V6, NOX4, and TGF␤1 signaling in IPF
ine and phosphothreonine on TGF␤1RI (lane 2, HAS2 ϩ Cont shRNA), which is blocked by CD44V6 shRNA (Fig. 7B, lane 3). Further, TGF␤1-induced up-regulation of phosphoserine and phosphothreonine (lane 4, TGF␤1 ϩ Cont shRNA) was also blocked by CD44V6 shRNA (Fig. 7B, lane 5). These results show that serine/threonine kinase(s) actively phosphorylates TGF␤RI when it is associated with CD44V6 and when HAS2 is overexpressed. These results also provide strong evidence that HA-mediated TGF␤RI kinase activity is CD44V6-dependent and that TGF␤1-stimulated TGF␤RI kinase activity involves CD44V6. Although the TGF␤RI is known to activate SMAD2 and SMAD3, the profibrotic effects of TGF␤1 signaling have been largely attributed to SMAD3 signaling (96). Fig. 7C shows that serine/threonine phosphorylation of SMAD3 is also upregulated by HAS2 overexpression and TGF␤1 treatments in HNLFb cultures, which is blocked by CD44V6 shRNA. Therefore, these results provide strong evidence that HA, the physiological ligand for CD44V6, has an important role in activating CD44V6-associated TGF␤RI kinase activity that is required for the onset of the SMAD3-mediated TGF␤1 response that promotes the progressive fibrosis in IPF.
CD44v6 enhances hepatocyte growth factor (HGF), VEGF, and EGF-dependent tyrosine kinase receptor (c-MET, VEGF receptor, and EGF receptor) activation of the downstream MAPK-signaling cascade through a positive-feedback loop (97)(98)(99). Therefore, we examined whether down-regulation of CD44V6 inhibits TGF␤RI/SMAD3 signaling and whether constitutive activation of TGF␤1 can recover production of CD44V6 in HNLFb cultures in which CD44V6 expression is depleted. RT-PCR results in Fig. 7D show that CD44V6 siRNA effectively inhibits up-regulation of CD44V6 by TGF␤1, and Western blots in Fig. 7E show that CD44V6 siRNA also effectively inhibits TGF␤RI phosphorylation and activation of p-SMAD3. Western blots showed that the amount of SMAD3 protein in this experiment was not altered (Fig. 7E). The results in Fig

CD44V6 regulates TGF␤1/SMAD3-mediated NOX4 induction and ROS production
The findings for Fig. 7 (A-E) provide support for a positivefeedback loop in which CD44V6-dependent activation of TGF␤RI/SMAD3 can result in synthesis of the NOX4 expression. Therefore, we determined the role of the SMAD3/NOX4 pathway in ROS generation and whether CD44V6 is required for initiating the TGF␤RI/SMAD3/NOX4-induced fibrogenic phenotype. Fig. 7G shows the results of analyses of dichlorofluorescein diacetate (H 2 DCFDA) fluorescence (a measurement of intracellular ROS) in TGF␤1-treated HNLFb cultures. A marked increase in fluorescence showed that intracellular ROS increased in HNLFbs treated with TGF␤1, which was significantly decreased by transfecting the cells with either SMAD3 shRNA, NOX4 shRNA, or CD44V6 shRNA before TGF␤1 treatment. Further, an optimal concentration of GKT137831 (1 M; the dose obtained from the dose-response curve of GKT137831 on NOX4 expression (data not shown)), a pharmacologic inhibitor of NOX1/NOX4, also inhibited ROS ϳ50% in the TGF␤1treated HNLFb cultures. These data support a role for CD44V6 signaling via TGF␤RI and SMAD3 in the induction and activation of the NOX4 response in TGF␤1-treated HNLFbs. In addition, the positive-feedback loop coupling Figure 7. CD44V6/TGF␤RI kinase and signaling events in lung fibrosis. A, Duolink assays were performed in IPFFbs and HNLFbs to analyze the association between TGF␤RI and CD44v6. Data in A (bottom) are expressed in relative fluorescence units adjusted for protein concentration (mean Ϯ S.E. (error bars); n ϭ 3; *, p Յ 0.005). B, HNLFbs were transfected with vector control, HAS2 cDNA, or TGF␤1 cDNA or pretransfected with control shRNA or CD44V6 shRNA either alone or followed by HAS2 or TGF␤1 overexpression. After 72 h, lysates were prepared and immunoprecipitated (IP) with anti-TGF␤RI antibody. The components in the immunoprecipitate were separated by SDS-PAGE and analyzed by Western blotting with anti-phosphoserine, anti-phosphothreonine, and anti-TGF␤RI antibodies. C, the lysates were immunoprecipitated with anti-Smad3 antibody, separated by SDS-PAGE, and analyzed by Western blotting with anti-phosphoserine, anti-phosphothreonine, and anti-Smad3 antibodies. D, real-time PCR analyses are shown for CD44v6 in HNLFbs transfected with control shRNA or CD44V6 shRNA, followed by incubation without (left) or with TGF␤1 for 24 h (right). ␤-Actin was used as a loading control. E, Western blots are shown for Smad3 phosphorylation, NOX4, and ␤-actin in lysates from cells harvested from the experiment in C. F, IPFFbs expressing control shRNA or two different V6 . Total RNAs were examined by real-time PCR analysis for CD44V6 mRNA expressed relative to GAPDH. G, HNLFbs were treated without or with control siRNA, SMAD3 siRNA, NOX4 siRNA, or CD44V6 siRNA (v6) for 24 h and then treated with TGF␤1 for 12 h. For comparison, HNLFbs were pretreated with 1 M GKT137831 (GKT; a NOX1/NOX4 inhibitor) alone for 8 h and then treated with TGF␤1 for 12 h. Intracellular ROS generation was detected by measuring H2DCFDA fluorescence. C-F show data representative of three experiments. The experimental data in B and G are from three sets of three independent experiments. Statistical analysis was with ANOVA; B and G, *, p Յ 0.005 versus the respective control group.

CD44V6, NOX4, and TGF␤1 signaling in IPF
CD44V6 and TGF␤RI activation shown in Fig. 7E can produce sustained SMAD3/NOX4 signaling, which can promote progressive fibrosis. This is supported by evidence that SMAD3/NOX4 signaling is required for lung myofibroblast activation and differentiation (16).

The CD44V6-NOX4 pathway induces TGF␤1-induced ECM contractility
Tissue repair is mediated by myofibroblasts through secretion of ECM proteins and by remodeling and contracting the extracellular matrix (92,100). Therefore, we investigated the role of the CD44V6-NOX4 pathway in contractile function of myofibroblasts by measuring the contractility of three-dimensional collagen matrices and by analyzing the expression of ␣-SMA, the cytoskeletal protein responsible for contractile actin stress fibers (51). Fig. 8 shows that TGF␤1-induced collagen gel contractility (A) and expression of ␣-SMA (B) were significantly inhibited by the GKT137831 (NOX4/NOX1 inhibitor), by knockdown of NOX4, by blocking CD44V6 with an inhibitory CD44V6 peptide (V6-PEP; see Fig. 10 for the peptide function), and by inhibiting intracellular ROS with catalase expression vector. Overall, these results indicate a crucial role for CD44V6-NOX4-dependent ROS activation in ECM con-tractile properties that occurs when myofibroblasts differentiate in response to TGF␤1.

Nox4/Ap-1-dependent mechanism of CD44v6 and Has2 expressions in primary lung fibroblasts
A previous report (76) provides evidence that NOX-dependent ROS regulates CD44 expression in atherogenesis. Therefore, we determined whether NOX4 has a critical role in the regulation of CD44v6 and HA expression in MNLFbs treated with TGF␤1. Fig. 9A shows that TGF␤1 treatment for 4 h resulted in a 2.5-fold increase in transcription of CD44v6 mRNA in wild-type MNLFbs, which increased to 7-fold by 8 h, whereas TGF␤1 had no significant effect in Nox4-null lung fibroblasts (Nox4-null MNLFbs). Further support for redox-dependent regulation of CD44v6 mRNA expression was shown by its significant inhibition in MNLFbs that had been pretransfected with NOX4 shRNA or pretreated with the Nox1/Nox4 inhibitor GKT137831 followed by TGF␤1 treatment (Fig. 9A).
Because recent studies from our laboratory and other show that (i) proinflammatory TGF␤1 up-regulates CD44V6 (1), (ii) TGF␤1 up-regulates NADPH oxidase (16), and (iii) NOX-derived ROS regulates Ap-1 (101), we investigated whether Nox4derived ROS, CD44v6, and Ap-1 activation are interrelated during lung fibrosis. Fig. 9B shows that Ap-1 DNA-binding activity increased significantly (ϳ60%) at 2 h and remained elevated through 12 h (ϳ300%) after TGF␤1 treatment in MNLFbs. In contrast, Nox4-null cells treated with TGF␤1 showed only a minimal response (Fig. 9B). To determine whether increased Ap-1 DNA-binding activity induced by NADPH oxidase is the mechanism for increased CD44v6 expression, MNLFbs, IPFFbs, and Nox4-null MNLFbs were co-transfected with  (71)). Fig. 9C shows that the luciferase activities in TGF␤1-treated MNLFbs and IPFFbs transfected with the WT CD44 Luc construct had significantly more luciferase activity than the same cells transfected with the AP-1-M construct. In contrast, transfection of Nox4-null MNLFbs, or IPFFbs that were transfected with NOX4 siRNA, with either WT CD44 Luc or the CD44 Luc (AP-1-M) construct showed no differences in luciferase activities ( Fig. 9C; the inset validates the NOX4 shRNA). The CD44 promoter regulates transcription of CD44s and variants, including the v6 exon. Because NOX4 shRNA down-regulated CD44v6 expression in both fibrogenic mouse lung fibroblasts (21dBLMFbs) and IPFFbs (Fig. 4, A and B), one of the interpretations of Fig. 9C is that lowering of Nox4 activity in Nox4-null MNLFbs down-regulates CD44v6 transcription. Because blocking CD44V6 inhibits CD44V6-containing isoforms (see Fig. 16 in our companion paper (71)) and because NOX4 regulates CD44V6 (Fig. 4, A and B), the experiment in Fig. 9C cannot rule out the down-regulations of transcriptions of all other CD44V6-containing isoforms.
ROS production in inflammatory atherosclerosis disease is largely mediated through the interaction of CD44 with HA (76). To address this, HAS2 mRNA in MNLFb and NOX4-null MNLFb cultures was measured after TGF␤1 treatment for various times by real-time PCR in quiescent cells. MNLFbs show  Figure 9. NOX4-Ap-1-dependent regulation of CD44V6 and of hyaluronan synthesis and degradation. A, real-time PCR analysis of CD44v6 expression in MNLFbs and Nox4 null MNLFbs treated with 2.5 ng/ml TGF␤1 are shown for the indicated times. Data are expressed as -fold change in gene expression relative to untreated controls. B, DNA-binding activity of Ap-1 was assayed in MNLFbs and Nox4-null MNLFbs after TGF␤1 treatment for the indicated times. Data are expressed as absorbance at 450 nm. C, Ap-1-dependent CD44 promoter activity was assayed in NLFbs and Nox4-null MNLFbs and in IPFFbs pretransfected with either control siRNA or NOX4 siRNA. Cells were transiently co-transfected with WT CD44-Luc or CD44-AP-1 mutant (M) plasmids and with a ␤-galactosidase control plasmid. After 48 h, cells were treated with vehicle or with 2.5 ng/ml TGF␤1 for 24 h, followed by luciferase and ␤-galactosidase activity measurements. Data represent -fold change in luciferase activity, normalized to ␤-galactosidase activity in TGF␤1-treated cells compared with their respective controls (mean Ϯ S.E. (error bars); n ϭ 3; *, p Ͻ 0.005). D, real-time PCR analysis of Has2 mRNA expression in MNLFbs and Nox4-null MNLFbs treated with 2.5 ng/ml TGF␤1 for times up to 12 h. Data represent -fold change in gene expression, normalized to ␤-actin and relative to untreated MNLFb control. E, MNLFbs and Nox4-null MNLFbs were treated with 2.5 ng/ml TGF␤1 or 50 g/ml Healon (1000-kDa HA), HA500, or HA150 for 12 h. Nox activity (ROS generation) was detected by measuring H2DCFDA fluorescence. F, MNLFbs, NOX4-null MNLFbs, and CD44-null MNLFbs were treated with 2.5 ng/ml TGF␤1 or HA150 alone for 12 h or pretreated with 2.5 ng/ml TGF␤1 for 12 h and then treated with HA150 for another 12 h. ROS generation was detected by measuring H 2 DCFDA fluorescence. G, MNLFbs and Nox4-null MNLFbs were treated with 2.5 ng/ml TGF␤1 or 50 g/ml Healon (1000 kDa), HA500, HA150 for 12 h or first treated with or without 12. very little expression of Has1 and Has3 (Fig. 9D). Therefore, Has2 is the major hyaluronan synthase in MNLFbs. Fig. 9D shows that Has2 mRNA expression increased continuously over 12 h with TGF␤1 in the MNLFbs but showed minimal response in the Nox4-null MNLFbs. Consistent with the increased CD44v6 expression by Nox4, this result indicates that Nox4 has a putative role in regulating Has2 expression and also provides evidence of involvement of HA interaction with CD44s and/or its variants. TGF␤1 also had a smaller effect on Has1 and Has3 expression in either cell type (Fig. 9D).
To determine the specificity of HA for ROS production, the MNLFbs and CD44-null MNLFbs were treated with Streptomyces hyaluronidase (control; heat-inactivated hyaluronidase) prior to HA treatment. Hyaluronidase removes extracellular HA and thus demonstrates the specificity of HA treatment. Fig.  9G shows that the hyaluronidase treatment abrogated the HA150-and HA500-increased ROS production in MNLFbs, whereas, in Nox4-null MNLFbs, these treatments had no effect. Hyaluronidases have been reported to be involved in the intracellular and extracellular catabolism of HA in a CD44-dependent manner (34). The hyaluronidase-treated cultures may have produced LMWHA species; thus, we cannot rule out the effect of LMWHA produced by hyaluronidase on the exogenous addition of LMWHA contributing to NOX activity, as seen in Fig. 9G. Further, TGF␤1 treatment also caused a timedependent increase in the mRNA expression of hyaluronidases Hyal3 and Hyal1 in MNLFbs, but not in Nox4-null cells (Fig.  9H). Hyal3 mRNA expression was 3.5-fold higher, and Hyal1 mRNA expression was 2.7-fold higher in MNLFbs than in Nox4-null cells within a short time period (at 8 h) after TGF␤1 treatment. TGF␤1 had no significant effect on Hyal2 expression in either cell type (Fig. 9H). Thus, these data suggest that Nox4 activation mediates TGF␤1-induced Has2 and hence HA levels in MNLFbs by regulating both Has2 (synthetic enzyme) and Hyal3/1 (degradation enzymes). Furthermore, ROS-regu-lated expression of Hyal3 and Hyal1 may lead to enhanced degradation of HA and generation of ROS, stimulating LMWHA. Thus, NOX4 activation increases HA synthesis, and LMWHA (HA150 and HA500) enhances Nox4-mediated intracellular ROS levels, suggesting a positive-feedback loop between HA and Nox4 activation, and this function of LMWHA requires CD44v6. Conversely, when HA150 and HA500 were treated with heat-inactivated hyaluronidase, this effect was reversed, indicating the specificity of LMWHA for regulating ROS production in a Nox4-dependent way (Fig. 9G).

Therapeutic targeting of CD44v6 is effective in fibrosis resolution
The co-receptor function of CD44v6 requires the molecule to be membrane-bound and to express the ectodomain of CD44v6 (see Figs. 15 and 16 in our companion paper (71)). Moreover, the CD44V6 shRNA/nanoparticle was well described in our previous paper (45) and was found to downregulate the TGF␤1-stimulated ␣-SMA and collagen-1 production in fibrogenic lung fibroblasts (1). In this study, we determined whether CD44v6 peptides block co-receptor function of the CD44v6 isoform.
CD44V6 peptides block the co-receptor function of the CD44v6 isoform-We prepared three peptides from murine and human CD44V6 sequences (human V6-PEP and mouse v6-PEP-2) (box in Fig. 10A) and human V6-PEP-1 (outside the box). Bioinformatic analyses indicate that the Lys and Glu (Fig.  10A, green KE) of the V6-PEP peptide can interact with Glu 119 and Ser 49 of the TGF␤RII activation site, and similar results have been reported for interaction of a TGF␤1 binding site in the TGF␤RII activation site (102). Moreover, V6-PEP has an Arg residue (Fig. 10A, green R), which interacts with the TGF␤RII by two hydrogen bonds, and this finding parallels a recent published study (102). The V6-PEP-1 peptide did not present any residue interacting in the same region, and for this reason, it was used as a control.
We tested whether these peptides would compete with endogenous CD44v6 and inhibit the activation of TGF␤1 with its receptor TGF␤R. The human V6-PEP effectively inhibited TGF␤1-dependent activation of p-SMAD in IPFFbs (Fig. 10B) and BLMFbs (Fig. 10E), whereas the human V6-PEP-1 did not. Each of the 3 amino acids (Lys, Glu, and Arg, indicated in Fig.  10A) within the V6-PEP sequence were mutated, and the mutated (M) peptides were unable to inhibit the TGF␤1-induced activation of p-SMAD3 in the IPFFb cultures (Fig. 10C). Fig. 10 (E and F) shows that the mouse v6-PEP-2 inhibited the TGF␤1 activation of p-SMAD3 in 21dBLMFbs cultures, and mutations of the Glu and Lys (indicated in Fig. 10A) did not (Fig. 10F).
Interestingly, the peptides are not species-specific, and the addition of the murine peptide (v6-PEP-2) as well as the human V6-PEP to murine 21dBLMFbs cultures inhibited the activation of p-SMAD3 (Fig. 10, E and F). Similarly, the murine v6-PEP-2 also inhibited p-SMAD3 activation significantly in IPFFbs (data not shown). Therefore, it is likely that both the human and murine homologous sequences confer the co-receptor function of CD44V6-containing isoforms for TGF␤1/ TGF␤RI/p-SMAD3 activation, and we used the human V6-PEP in subsequent experiments. Fig. 10D shows that a monoclonal antibody to TGF␤1 interacts with the V6-PEP peptide bound to peripheral blood mononuclear cell (PBMC) cultures equivalent to its binding to TGF␤1 bound to the isolated PBMC cultures. This indicates that the v6-PEP peptide mimics TGF␤1, confirming that the TGF␤1/TGF␤R interaction on the isolated PBMCs has equally recognized both TGF␤1 and the V6-PEP peptide.
The V6-PEP/nanoparticle and the CD44V6 shRNA (V6 shRNA)/nanoparticle are non-cytotoxic and mediate an antiinflammatory response-To demonstrate that the V6-PEP/ nanoparticle and the V6 shRNA/nanoparticle mimic TGF␤1 without cytotoxic action, we performed the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays using PBMC cultures. Fig. 11A shows that the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle tested at 1, 10, and 100 g/ml concentrations did not affect PBMC viability and presented no significant differences from controls. To confirm whether the V6-PEP/nanoparticle and V6 shRNA/nanoparticle have the same ability as TGF␤1 to modulate an immune response, we stimulated PBMCs with either LPS, our V6 shRNA/nanoparticles, or V6-PEP/nanoparticles and measured TNF␣ and IL-10 production by an ELISA. The V6-PEP/nano-particle and the V6 shRNA/nanoparticle were not able to induce TNF␣ (Fig. 11B) or IL-10 ( Fig. 11C) production in the absence of inflammatory stimulus. PBMCs pretreated with the different concentrations of the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle followed by LPS stimulation for 24 h significantly decreased TNF␣ production when compared with LPS-treated cells (Fig. 11D). Interestingly, all concentrations of both the V6-PEP/nanoparticle and the V6 shRNA/ nanoparticle significantly increased anti-inflammatory IL-10 production when compared with controls (Fig. 11E). Furthermore, the efficiency of both nanoparticles in down-modulating TNF␣ and up-regulating IL-10 was sustained after a 48-h stimulus of PBMCs before LPS treatment for 24 h (data not shown). Overall, the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle are non-toxic, and they significantly reduced inflammatory responses. Thus, the peptide and shRNA against CD44V6 could down-regulate pro-inflammatory TNF␣ and up-regulate anti-inflammatory IL-10 production during inflammatory stimuli, a profile that is required in the treatment of inflammatory diseases (103).
The V6-PEP/nanoparticle and the V6 shRNA/nanoparticle inhibit the fibrogenic lung fibroblasts' ability to promote migration of mononuclear leukocytes across endothelial cell monolayers-Perivascular infiltrates of inflammatory cells are a hallmark of injured tissue in tissue fibrosis. Thus, we deter-

CD44V6, NOX4, and TGF␤1 signaling in IPF
mined whether the V6-PEP/nanoparticle and the V6 shRNA/ nanoparticle can inhibit fibrogenic lung fibroblast-mediated mononuclear leukocyte migration across endothelial cell (EC) monolayer cultures. A three-cell co-culture system, together with complementary conditioned medium transfer experiments, was used to determine whether fibrogenic lung fibroblast cultures promote migration of mononuclear leukocytes across EC (C166 cells) monolayers. Fig. 12A shows that co-culture of ECs with 21dBLMFbs in the lower Boyden chamber increased leukocyte (PBMC) migration through the EC layer by ϳ2.3-fold, which increased to ϳ3.5-fold when the 21dBLMFbs were treated with the inflammatory stimulant LPS, compared with migration through the EC layer in the absence of the 21dBMLFbs in the lower chamber. In contrast, EC co-culture with normal MNLFbs showed only a minimal increase in leukocyte migration. Further, co-culture of ECs treated with LPS with untreated 21dBMLFbs did not increase leukocyte migra-tion (Fig. 12A), possibly, at least in part, due to the greatly increased adhesion of the leukocytes to the LPS-treated ECs (Fig. 12B). No increase of leukocyte adhesion was observed with the other co-culture experiments (Fig. 12B), where leukocytes migrate substantially through the EC layer (Fig. 12A). Next, we showed (Fig. 12C) that treatment of the 21dBLMFbs with LPS followed by treatment with the V6-PEP/nanoparticle or with the V6 shRNA/nanoparticle effectively prevents the ability of the LPS treatment to stimulate PBMC migration through the EC layer. Similarly, treatment of human IPFFbs with LPS and then with the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle also prevents the subsequent human U937 (leukocytes) migration response on an HUVEC monolayer (Fig. 12D). These results indicate that the fibrogenic lung fibroblasts promote transendothelial leukocyte migration by altering the properties of the EC layer.
To investigate this further, a series of conditioned medium (CM) transfer experiments were done using IPFFb CM to mea- Figure 11. The CD44V6 shRNA (V6 shRNA) and the V6-PEP/nanoparticles are non-cytotoxic and mediate anti-inflammatory responses. A, cytotoxicity analysis (by MTT assay) results are shown for V6 shRNA/nanoparticles and V6-PEP/nanoparticles in PBMCs (leukocytes). The V6 shRNA/nanoparticle (pSico-CD44V6 shRNA/Tf-PEG-PEI plus FSP-1-Cre/Tf-PEG-PEI) and v6-PEP/nanoparticle (1-100 g/ml) showed no statistical difference when compared with control (PBMCs with control peptide treatment; the data of other control (PBMCs with control shRNA/nanoparticle (pSico-scrambled shRNA/Tf-PEG-PEI plus FSP-1-Cre/ Tf-PEG-PEI) treatment) show the same cytotoxicity as PBMCs with control peptide treatment (data not shown)), indicating that the tested concentrations did not have a cytotoxic effect. B-E, analyses of cytokine production by PBMCs after stimulation for 24 h. Analyses of pro-inflammatory TNF-␣ (B) and regulatory IL-10 (C) cytokines released in the absence of inflammatory stimulus are shown. Analysis of TNF-␣ (D) and IL-10 (E) cytokines released after PBMC stimulation with 1 g/ml LPS are shown. PBMC cultures pretreated with the V6-PEP/nanoparticle or with V6 shRNA/nanoparticle at 1 g/ml were able to decrease TNF-␣ production (*, p Ͻ 0.001). PBMC cultures pretreated with the V6-PEP/nanoparticle or with V6 shRNA/nanoparticle followed by stimulation with 1 g/ml LPS resulted in a significant increase in IL-10 production when compared with control (Ϯ S.D.; n ϭ 3; *, p Յ 0.001).
sure U937 monocyte migration through the HUVEC layer. Fig.  12E shows that IPFFb CM substantially increased U937 leukocyte migration into the lower chamber. Further experiments investigated whether transmigration of leukocytes might involve chemokine mediators. IL-8, a C-X-C chemokine, has been implicated as a promoter of leukocyte extravasation in fibrosis (104), whereas the C-C class of chemokines predominantly affects the migration of mononuclear leukocytes (105) and includes MCP-1 (monocyte chemoattractant protein-1). Fig. 12E shows that IL-8 consistently induced a small increase in U937 cell migration compared with control wells, whereas MCP-1 induced transendothelial migration of U937 cells into the lower chamber equivalent to IPFFb CM alone. Further, preincubation of IPFFb CM with anti-MCP-1 antibody, with anti-CD44v6 antibody, with V6-PEP/nanoparticles, or with V6 shRNA/nanoparticles reduced U937 migration back to the level inserts were counted by hemocytometer. 21dBLMFbs significantly increased migration of PBMCs through the EC layer, which was further increased by cytokine activation of 21dBLMFbs (1 g/ml LPS for 8 h) prior to co-culture with ECs. *, p Ͻ 0.01, statistically significantly greater than migration across resting EC layer. B, PBMCs attached to the insert 16 h after migration were measured by colorimetry (see "Experimental procedures"). Activation of the EC cell layer with cytokines reduced the number of PBMCs migrating across the EC but increased the PBMC binding to the EC monolayer. C and D, pretreatment of 21dBLMFbs with 1 g/ml LPS for 8 h followed by treatment with 1 g/ml V6-PEP/nanoparticle or 1 g/ml V6 shRNA/nanoparticle reduced PBMC migration through the control EC monolayer (C) and also reduced U937 (leukocyte) cell migration on the HUVEC monolayer (D). *, p Ͻ 0.005. E, promotion of U937 (leukocyte) cell migration by CM derived from IPFFbs, by recombinant monocyte chemoattractant protein-1 (MCP-1), and by IL-8 are shown. ϳ250 ϫ 10 3 U937 cells were put onto confluent HUVEC monolayers that were treated with human recombinant MCP-1, with CM from IPFFb cultures alone or cultures that were treated with IL-8 (50 ng/ml), with 100 g of each of the indicated nanoparticles, or with a 500 g/ml concentration of either anti-MCP-1 or anti-CD44v6 for 12 h. U937 cell numbers in the lower chamber were counted in triplicate wells 16 h after the addition of the reagents and the CMs to the U937 cells and HUVECs on the inserts. of IL-8 and near the level seen in control wells (Fig. 12E). These results indicate that release or activation of the MCP-1 C-C chemokine by fibrogenic fibroblasts provides another likely mechanism by which fibrogenic lung fibroblasts can promote transendothelial leukocyte migration and that this chemokine is at least partly regulated by CD44v6. This is further supported by evidence that hyaluronan binding with CD44 regulates MCP-1 in renal tubular epithelial cells (106). These data suggest that fibrogenic lung fibroblasts promote leukocyte migration across endothelial cell monolayers in tissue culture via a CD44v6-dependent mechanism that possibly includes MCP-1, which is likely to be relevant to the perivascular mononuclear leukocyte infiltrates characteristic of early fibrotic lesions (107).
Role of the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle in bleomycin-induced lung fibrosis-Previous studies have shown that intratracheal instillation of bleomycin induces epithelium injury that leads to fibrosis, which peaks 14 -28 days post-injury followed by gradual resolution (108) (see timeline in Fig. 13A). Although CD44v6 was induced in mice starting at 3-7 days after bleomycin injury, its expression remained elevated in the fibrogenic phase (Fig. 2 in our companion paper (71)). Using our non-toxic and non-inflammatory shRNA/ nanoparticle delivery method (45, 109), we evaluated the efficacy of targeting CD44v6 in the bleomycin fibrotic lung model by intratracheal delivery of v6 shRNA/nanoparticles and v6-PEP/nanoparticles starting on day 2 and every 2 days for 30 days after bleomycin injury (details of nanoparticle delivery are discussed under "Experimental procedures"). The preparation of nanoparticles and the delivery was done following our published method (45,109). The nanoparticles contained 1) V6 shRNA/nanoparticle (genetic approach) and 2) the V6-PEP/ nanoparticle (a small-molecule approach). Fig. 13B shows that both the CD44v6 knockdown by the V6 shRNA/nanoparticle and blocking by the V6-PEP/nanoparticle restored the capacity for resolving fibrosis by day 30 as determined by less Masson's trichrome blue staining for collagen (Fig. 9B), by significantly reducing CD44v6 and Nox4 mRNA and protein expression (Fig.  13, C and D), and by suppressing hydroxyproline (collagen) content (Fig. 13E). Fig. 13F shows that lung weights for the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle treatments were significantly lower compared with control shRNA/ nanoparticle treatment. Fig. 13G shows that the mice given the V6-PEP/nanoparticle or the V6 shRNA/nanoparticle had less body weight loss and returned body weights to control levels by 30 days, in contrast to the mice receiving the control shRNA/ nanoparticles. Fig. 13H shows that the survival of mice treated with the control shRNA/nanoparticle or control peptide/nanoparticle decreased during 21-45 days after treatment, whereas , or V6-PEP/nanoparticle were administered every other day from day 2 to day 30 by intratracheal delivery to the lungs of young mice during the onset of inflammation by bleomycin injury and then left untreated for another 15 days (see "Experimental procedures" for a detailed description of the shRNA/nanoparticle preparation and delivery method). Lung tissue was harvested at the indicated times up to 45 days after injury. The time course of fibrosis induction, treatment schedule, and resolution of fibrosis are shown. Nanoparticle preparation and delivery method was validated by our group previously in an intestinal/colon cancer murine model (45,153). B, fibrosis was assessed by Masson's trichrome blue staining for collagen in sections of lungs isolated at day 24. Fibroblasts were isolated from lungs at 0, 7, 14, 21, and 45 days after bleomycin treatment in mice treated intratracheally with or without CD44V6 shRNA/nanoparticle, control shRNA/ nanoparticle, or V6-PEP/nanoparticle and cultured ex vivo. C, total RNA was isolated from these fibroblasts, and real-time PCR analyses were done for CD44v6, Nox4, and Gapdh mRNA expression. D, whole-cell lysates were prepared from the isolated fibroblasts from the indicated days after bleomycin injury. Immunoblotting analyses were done for CD44v6, Nox4, and GAPDH protein expressions. E, quantitative hydroxyproline collagen assays were done for tissue samples from lungs collected at 0, 7, 14, 21, and 45 days after bleomycin injury in mice treated with or without CD44V6 shRNA/nanoparticle or control shRNA/ nanoparticle. the mice treated with the V6-PEP/nanoparticle or the V6 shRNA/nanoparticle survived.

Discussion
IPF is a chronic progressive fibrosis interstitial lung disease. The underlying etiology in IPF is generally unknown. Currently, IPF affects about 100,000 people in the United States, with a high 5-year mortality among those with severe progressive disease (110,111). Pirfenidone and nintedanib were each shown to reduce the average decline in lung function in randomized clinical trials (112). Both drugs have demonstrated efficacy in phase III clinical trials by retarding the rate of progression of IPF, but neither drug appears to be able to completely arrest disease progression (112). Thus, lung transplantation remains the only viable intervention in end-stage disease (111). Patients with IPF are recognized as a high unmet medical need with increasing prevalence within the aging population. Understanding the key cellular processes, including proliferation, acquisition of a contractile phenotype, extracellular matrix protein synthesis and secretion, and cell fate decisions, including cell survival and apoptosis, provides an opportunity to explore potential novel therapeutic approaches. One of the major profibrotic cytokines promoting fibrogenesis is TGF␤1. The pathologic changes in IPF (excessive accumulation of ECM and remodeling of the lung architecture) result when the balance between fibroblast proliferation and apoptosis is shifted toward accelerated proliferation and/or slowed apoptosis of fibroblasts, the primary ECM producers (113). A number of studies, including our own, have demonstrated enhanced autocrine TGF␤1 signaling in myofibroblasts, the principal cell type associated with the pathogenesis of disease (1, 26, 114 -117).
Our study (7) demonstrated that the HA receptor CD44V6 induces autocrine TGF␤1 signaling. However, the signal transduction cascades by which CD44V6 induces TGF␤1 responses were not elucidated.
NOX4 is a component of the TGF␤1 signaling pathway, with NOX4-derived H 2 O 2 acting as a signaling molecule responsible for activating downstream cell survival signaling pathways, including the AKT and MAPK pathways (118,119). Studies have demonstrated that TGF␤1 is responsible for the induction in NOX4 expression by pulmonary fibroblasts surrounding pulmonary vessels, suggesting that NOX4 is a component of the TGF␤1 signaling to generate extracellular H 2 O 2 (16,120,121). In addition, NOX4 has been implicated in the profibrotic responses in lung myofibroblasts (16,122). Despite its critical role in modulating profibrotic responses stimulated by TGF␤1 and NOX4, the regulation of CD44V6 expression and activity, as well as the mechanisms underlying their mutual regulation in fibrotic diseases, has not been reported.
The present results reveal the following. 1) Levels of CD44V6 and NOX4 are significantly elevated in IPF lung biopsies and in lung biopsies from bleomycin-induced lung injury (Figs. 1A,  3A, and 5A). Moreover, the expressions of CD44V6 and NOX4 in explanted normal lung fibroblasts were markedly up-regulated by TGF␤1 (Figs. 1B, 3B, 4 (A and B), and 5 (B, D, and F)). 2) CD44V6 regulates TGF␤1-induced NOX4 expression in ex vivo normal lung fibroblasts (Figs. 6C and 7E), and stimulation of NOX4 by CD44V6 involved the activation of TGF␤RI kinase and canonical TGF␤RI-independent p-SMAD3 serine threonine kinase activation, which then directly stimulated NOX4 expression (Fig. 7, B, C, and E). 3) Nox4 also enhanced TGF␤1induced CD44v6 expression via Ap-1 activation (Fig. 9, B and C). 4) Nox4 is directly involved in regulating the increased synthesis of Has2 in TGF␤1-treated murine lung fibroblasts (Fig.  9D). 5) TGF␤1 up-regulates Hyal3 and Hyal1 hyaluronidases predominantly, suggesting that degradation of HA in TGF␤1treated lung fibroblasts is also regulated by Nox4 (Fig. 9H). 6) Interaction of LMWHA with the increased CD44v6 expression induced in TGF␤1-treated lung fibroblasts that depends on Nox4 has an important role in cell proliferation, differentiation, and ECM protein and contractile protein expressions in the ex vivo murine and human lung fibroblasts (Figs. 4 (A and B), 5B, 6A, and 8). 8) NOX4-increased ROS production activated AKT, MMP2, and MMP9 expressions, which mediated the cell invasion of lung fibroblasts (Fig. 6, B and D). 9) CD44v6-and HA-dependent Nox activity and intracellular ROS expressions are significantly up-regulated in the ex vivo wild-type MNLFbs compared with Nox4-null MNLFbs and CD44v6-null MNLFbs (Fig. 9, E and F). 10) Irrespective of the mechanism of co-receptor function, V6-PEP mimicked TGF␤1 by binding to a TGF␤1 monoclonal antibody, confirming that the PBMC cell surface TGF␤R is also equally recognized by V6-PEP and TGF␤1 (Fig.  10D), and inhibited co-receptor function of CD44v6 through inhibition of p-SMAD3 activation (Fig. 10, B, C, E, and F). 11) Both V6-PEP and co-receptor are not toxic (Fig. 11A) and are anti-inflammatory by inhibiting TNF␣ production (Fig. 11, B  and D), by increasing LPS-induced IL-10 production (Fig. 11E), and by abrogating transendothelial migration of leukocytes via fibrogenic lung myofibroblasts (Fig. 12, A, C, and E). 12) Inflammatory mediators of fibrogenic fibroblasts induce leukocyte migration through endothelial cell layers, and this function is mediated by CD44v6 and MCP-1, indicating that in addition to their role in ECM homeostasis, fibroblasts may influence inflammatory disease processes through an effect on CD44v6mediated leukocyte trafficking, which may involve MCP-1. 13) GKT137831 (1 M), a pharmacological inhibitor of NOX1/ NOX4-based NADPH oxidases, not only inhibits TGF␤1-induced H 2 O 2 but also attenuates TGF␤1-induced myofibroblast differentiation measured by collagen gel contraction (Fig. 8). 14) In the bleomycin-challenged mouse model of pulmonary fibrosis, the V6-PEP/nanoparticle and the V6 shRNA/nanoparticle strategies targeted CD44V6 induction and TGF␤1 activity, respectively, and suppressed fibrosis by inhibiting gene transcription of profibrotic and proinflammatory mediators and of ECM proteins in the lungs.
HA is synthesized by three HA synthases (33); however, HAS2 is the major HA-synthesizing enzyme in IPF, based on two lines of evidence. First, targeted overexpression of Has2 by myofibroblasts produced an aggressive phenotype, leading to severe lung fibrosis and death after bleomycin-induced injury (19). Second, conditional deletion of Has2 in mesenchymal cells abrogated the invasive fibroblast phenotype, impeded myofibroblast accumulation, and inhibited the development of lung fibrosis (19). Our results indicate that increased Has2 expression and activity in TGF␤1-treated lung fibroblasts is dependent on NOX4 activation (Fig. 9D). Furthermore, our data indi-

CD44V6, NOX4, and TGF␤1 signaling in IPF
cating that Nox4 regulates Hyal3 and Hyal1 expression in MNLFbs treated with TGF␤1 are significant because increased HA production during pathological conditions depends on the turnover of HA (i.e. the sum of its synthesis and catabolism). This observation is substantiated by the published studies indicating that Hyal3-null mice do not show HA accumulation (123) and Hyal3 contributes to HA metabolism by augmenting the activity of Hyal1 (124). Hyal3 is also enhanced by cytokines in chondrocytes (125) and also in vascular smooth-muscle cells in atherosclerotic disease (90). In a recently proposed model, the breakdown of HA is initiated extracellularly by HYAL3 and then continued intracellularly by HYAL1 (126). In addition, HA can also be degraded into smaller fragments in a non-enzymatic way by exposure to ROS intermediates (127), which is believed to be an important mechanism for generating HA fragments at sites of inflammation (128). It is generally accepted that free radicals, especially the highly reactive hydroxyl radical, have an important role in the degradation process of HA (127)(128)(129)(130)(131)(132), but the direct action of ROS on HA turnover and degradation has received scant attention to date.
Our data show that the increase in TGF␤1-induced NOX4 activity by HA500 and HA150 (Fig. 9, E and G) is due to enhanced CD44v6 expression ( Fig. 7D and Figs. 15 and 16 in our companion paper (71)) caused by TGF␤1 pretreatment. We cannot rule out the possibility that HA regulates TGF␤1-induced NOX4 activity through binding to CD44v6, because HA regulates Rac1 GTPase (95), which is a regulatory subunit of NADPH oxidase. We found that CD44-null MNLFbs and Nox4-null MNLFbs suppressed ROS production (Fig. 9F), and this finding is supported by the studies in which HA-CD44 interaction has been reported to induce ROS production (95), and down-regulation of p47 phox or p67 phox decreased ROS levels (95). Interestingly, our studies indicate that Nox4 regulates CD44v6 gene expression (Fig. 9C); Nox4 increases Has2 expression (Fig. 9D), which is known to produce HA; and LMWHA enhances Nox4/ROS activity (Fig. 9E), suggesting the presence of a positive-feedback loop between HA and NOX4 activation. These data support a key role for NOX4 and HA in the pathogenesis of lung fibroblasts in response to TGF␤1. Our results thus suggest that the NOX4-dependent expression of CD44v6 and HA in the fibrogenic lung fibroblasts and ROS-dependent HA degradation and release of LMWHA lead to increased interaction of CD44v6 and LMWHA and induction of pro-inflammatory cytokine expression in the fibroblasts.
Increased myofibroblast differentiation, invasion, proliferation, and synthesis of pro-fibrogenic molecules, including NOX4, CD44v6, and ECM protein (collagen-1 and fibronectin), as seen in Figs. 4A, 5 (B-G), and 6 (A-D) are important events in the development of fibrogenic processes. Serum-stimulated wild-type IPFFbs proliferate more rapidly than NOX4-or CD44V6-depleted IPFFbs (Fig. 6A). Our data also indicate that the importance of NOX4-dependent CD44V6 expression in pulmonary fibrosis is supported by our findings that CD44V6 shRNA significantly decreased TGF␤1-induced ␣-SMA, fibronectin, collagen-1, and NOX4 expressions, endothelial cell layer invasion, and cell proliferation and differentiation of IPFFbs and TGF␤1-stimulated HNLFb (Figs. 5 (B-G), 6 (A-D), and 8). We also found that soluble CD44V6 levels in the plasma of 21-day bleomycin-treated mice were increased compared with PBS-treated mice (data not shown), which further indicates that activation of NOX4 in fibrogenic settings releases soluble CD44V6 by proteolytic cleavage of membrane-anchored CD44V6. This may well be from the increased high levels of matrix metalloproteinases in IPFFbs (Fig. 6B).
Similar to a feedback loop between NOX4 and HA (Fig. 9, D and E), our studies further indicate that CD44V6 regulates NOX4 expression and function through TGF␤1-induced p-SMAD3 activation (Fig. 7, B and C). NOX4 also enhanced TGF␤1-induced CD44v6 expression via Ap-1 activation (Fig. 9, B and C), and CD44V6 regulates NOX4 expression and activity (Figs. 6C and 9 (E and G)), suggesting another positive-feedback loop between CD44V6 and NOX4 activation. Taken together with the previous demonstration (1) that augmented CD44V6 expression in interstitial lung disease fibroblasts sensitized them to the profibrotic effects of TGF␤1, the present results provide previously unreported insight into the mechanism of fibrosis in IPF and indicate a fundamental role of HA-CD44V6mediated NOX4/ROS in mediating this process. In this context, it is noteworthy that pharmacologic inhibition of NOX1/ NOX4 in TGF␤1-treated lung fibroblasts not only attenuated ROS generation but also decreased fibroblast differentiation to myofibroblasts (Figs. 7G and 8).
To explore the role of CD44V6 in fibrotic responses to injury of the mammalian lung, we used a murine model of acute lung injury. It has been reported that histological hallmarks, such as intra-alveolar buds, mural incorporation of collagen, obliteration of the alveolar space, and epithelial injury with subsequent mesenchymal cell activation and fibrosis, are present in bleomycin-treated animals similar to IPF patients (133). This observation has led to the assumption that bleomycin reproduces typical features of the human disease, including TGF␤1 up-regulation and activation, and myofibroblast differentiation and activation are recapitulated in this animal model. Therefore, the use of this model has become very popular. Further, the bleomycin model has the advantage that it is quite easy to perform and is widely accessible and reproducible, therefore fulfilling important criteria expected from a good animal model (16,133). We first showed that CD44v6 expression is induced during the fibrogenic phase of bleomycin-induced lung injury in a time-dependent manner, increasing from day 3 up to day 28 ( Fig. 2 in our companion paper (71) and Fig. 13 (C and D)), whereas NOX4 expression was induced from day 7 up to day 28 (Fig. 13, C and D), supporting a functional relationship between Nox4 and CD44v6 expression, myofibroblast activation, and fibrosis after lung injury. In contrast, expression of the Nox2 isoform, which is predominantly expressed in phagocytic cells, was increased on day 14 and returned to baseline levels at day 45, when inflammatory responses through synthesis of Nox4 had subsided (Fig. 3B).
We also examined the effects of targeted suppression of CD44v6 induction with in vivo CD44v6 shRNA/nanoparticle and v6-PEP/nanoparticle treatments compared with control shRNA/nanoparticles in the bleomycin animal model of lung injury and fibrosis. We found 1) that these nanoparticles have no off-target effects (i.e. they produce IFN␣ neither in in vitro (see Fig. 12H in our companion paper (71)) nor in in vivo set-tings (Fig. 2, C and D)); 2) that they are non-toxic (Fig. 11A), promote anti-inflammatory responses (Fig. 11, B and D), and abrogate fibrogenic lung fibroblast-mediated leukocyte migration across an endothelial cell layer (Fig. 12, C and D); and 3) that these antagonists targeting CD44V6 up-regulate IL-10 production (Fig. 11E) during inflammatory stimuli, a profile that is required in the treatment of inflammatory diseases for immune suppression (103,134).
These findings in Figs. 10 -12 explain the reasons that led us to select these novel lung fibroblast-specific CD44V6 shRNA/ nanoparticles through FSP-1-Cre/nanoparticle-mediated transactivation of the V6-PEP/nanoparticle and the V6 shRNA/ nanoparticle for intratracheal delivery in the bleomycin-treated mouse model. CD44V6 knockdown by lung fibroblast-specific V6 shRNA/nanoparticle and by competing for endogenous CD44v6 co-receptor function by v6-PEP/nanoparticle mediated a marked antifibrotic effect, as determined by Masson's trichrome staining for collagen (Fig. 13B), by biochemical analyses of hydroxyproline content (Fig. 13E) in whole lung homogenates, by inhibiting increased lung weight (Fig. 13F), by restoring body weight (Fig. 13G), and by mouse survival (Fig.  13H).
Our data indicating that lung fibroblast-specific genetic inhibition of CD44V6 markedly decreases NOX4 expression in the lung fibroblasts (Fig. 13, C and D) validate our hypothesis that HA-CD44V6 activation promotes fibrosis via NOX4-dependent processes. To our knowledge, this is the first study to show that a specific peptide inhibitor of CD44V6 activity significantly decreases profibrogenic processes in bleomycin-induced lung fibrosis in mice. This inhibitor may have therapeutic potential because it inhibited Nox4, Has2 expression, and oxidative stress in lung mesenchymal cells in pulmonary fibrosis.
In summary, our data provide insight into the molecular mechanisms by which CD44V6-mediated NOX4 activation promotes fibrogenic lesion formation. NOX4 mediates a TGF␤1-induced HA synthesis by regulating both HA synthetic and degradation enzymes, and a positive-feedback loop between HA and NOX4 increases oxidative stress and enhances fibrosis (model in Fig. 14). On the other hand, CD44v6 regulates NOX4 expression and function via TGF␤RI/ p-SMAD3 activation, and NOX4 regulates CD44v6 gene transcription via Ap-1 activation. Further, another positive-feedback loop between CD44v6 and NOX4 increases oxidative stress and enhances fibrosis. Interaction between increased CD44v6 and HA resulting from NOX4 activation in profibrogenic background induces cytokine and ROS production, which promote pulmonary fibrosis. Taken together, these results demonstrate that the CD44V6 is itself a target of TGF␤1 signaling in lung fibroblasts. Enhanced CD44V6 expression probably contributes to NOX4-mediated persistence and progression of the fibrotic response.
As discussed above, we evaluated the role of CD44v6 in mediating the fibrogenic response in several ways. However, CD44null mice did not show significant protection from the development of pulmonary fibrosis (19). Compensation of proteins by members of the same family has been widely proposed to explain the lack of phenotype of several knock-out mice. However, the effect on protection from the development of pulmo-nary fibrosis is more pronounced when the ␣-Sma-Has2 transgenic mice are bred with the CD44-null mice (19). However, our results indicate that progressive fibrosis requires the generation of an invasive myofibroblast phenotype that requires TGF␤1 where NOX4, HA, and CD44V6 are critical downstream components of the TGF␤1-induced fibrogenic response (Figs. 7 (B-G) and 9 (A, B, D, F, and G)). Moreover, NOX4 and CD44V6 regulate MMP2 and MMP9 for the emergence of the invasive phenotype (Fig. 6, B and D). Our studies also indicate that fibrogenic lung fibroblasts modulate leukocyte-endothelial cell interactions in ways that facilitate migration of leukocytes across EC layers, at least partly via a CD44V6-dependent mechanism involving MCP-1 activation. These results provide further support for the view that, in addition to their role in extracellular matrix homeostasis, fibroblasts may influence inflammatory disease processes through an effect on leukocyte trafficking. The findings that coordinated gene expression with up-regulation of matrix-degrading enzymes (MMPs), inflammatory soluble mediators like MCP-1, as well as fibroblast activation through CD44V6-NOX4 signaling occur in both mouse and human pulmonary fibrosis, suggesting that the approach to abrogate CD44V6 could be used to identify therapeutic targets. The proof-of-principle experiments suggest that the non-toxic and anti-inflammatory V6-PEP/nanoparticle or CD44V6 shRNA/nanoparticle have no off-target effects and attenuated lung fibrosis in mice in vivo at the time of bleomycin injury. Therefore these proof-of-principle studies Our studies indicate that NOX4 increases HAS2 expression, and LMWHA enhances NOX4/ROS activity, which provides evidence for a positive-feedback loop between HA and NOX4 activation. ROS then induces cell invasion through the Akt 3 MMP pathway. Our study also provides evidence that in response to lung injury, CD44V6 induces NOX4 through regulation of TGF␤1/SMAD3 signaling in myofibroblasts, and NOX4 also regulates HAS2 production as well as AP-1-mediated CD44V6 expression. This suggests that the CD44v6 variant and HA are up-regulated by TGF␤1/SMAD3 signaling through a feedback loop requiring the presence of NOX4. These studies thus show the involvement of CD44V6 in TGF␤-induced NOX4 expression, myofibroblast activation, and profibrotic responses. (Figs. 11-13) represent an innovative therapeutic strategy to control tissue fibrosis. Fig. 14) A network consisting of cytokines, oxygen radicals derived from inflammatory cells, closely controls activation of several ECM components to induce the progression of fibrosis. The activated inflammatory cells that accumulate in the lower airways release harmful amounts of ROS, which results in lung injury and proliferation of fibroblasts in alveolar walls (135). In this study, we show that bleomycin-induced mouse lung fibroblasts express the fibrosis-associated protein CD44V6 (1), which can facilitate presentation of TGF␤1/p-SMAD3 to its receptor TGF␤RI for induction of NOX4 and associated impaired oxidant-antioxidant balance during the fibrogenic phase. NOX4 also enhanced TGF␤1-induced CD44V6 expression via AP-1 activation and HAS2 production. Furthermore, we demonstrate that CD44v6 promotes the feedback up-regulation of NOX4 through TGF␤1-TGF␤RI signaling. Similarly, HA and NOX4 form a feedback loop to up-regulate ROS, which induces cell invasion through a ROS/AKT/MMP2/9 pathway. Because bleomycin is widely used in experimental models of human diseases resembling pulmonary fibrosis, in the present study, we investigated the possible protective effect of inhibiting CD44V6 against bleomycin-induced oxidative injury in the mouse lung. Our results together with biochemical and signaling data agree with this hypothesis, because the bleomycininduced increase in fibrotic activity, as assessed by the lung collagen content, is also reduced by the lung fibroblast-specific release of CD44V6 shRNA/nanoparticle and the V6-peptide/ nanoparticle treatments. In conclusion, the findings of the present study demonstrated for the first time that blocking CD44V6 has an additional protective effect on the source(s) of inflammation that induce production of NOX4 with subsequent ROS generation and deposition of pathological extracellular matrix components that result in pulmonary fibrosis .
Hyaluronidase isolated from Streptomyces hyaluronlyticus, which contains endotoxin, was from EMD Millipore (catalog no. 389561) and was used to digest hyaluronan. The hyaluronidase was removed by digesting with Proteinase K followed by denaturation at 95°C for 5 min. Endotoxin was removed by extracting twice with 1% Triton X-114. The aqueous phase contains the degraded hyaluronan (136). The endotoxin level was measured by an end-point chromogenic limulus amebocyte lysate assay.
Hyaluronan from Hyalose (Oklahoma City, OK) was prepared by enzymatic synthesis of specific molecular mass and was free from endotoxin. Pharmaceutical-grade hyaluronan from Life Core Biomedical (Chaska, MN) was also used.

Management of animals, human lung samples, and lung fibroblasts
Six-week-old mice (C57BL/6 strain) were obtained from the Jackson Laboratories. Mice (0.05 units/20 g of animal) were anesthetized with isoflurane. Intratracheal administration of bleomycin (0.025 units in PBS) was used to induce lung injury as described previously (22). All animal care and experimentation were done in accordance with the institutional animal care and use committee protocol (AR 3220) approved by the Medical University of South Carolina according to the rules of the National Institutes of Health. Lung tissues at 21 days after PBS or bleomycin instillations were perfused with z-fix (Anatech Ltd., Battle Creek, MI) and processed for paraffin sections. Fibroblasts from control mouse lungs (NMLFbs) and from 14and 21-day bleomycin-injured mouse lungs (14dBLMFbs and 21dBLMFbs) were isolated as described in our companion paper ( Fig. 1 in Ref. 71).

Cell culture
HNLFbs and lung fibroblasts from IPF patients were obtained from Dr. Carol Feghali-Bostwick (Medical University of South Carolina). Briefly, lung tissues were diced (ϳ0.5 ϫ 0.5-mm pieces) and cultured in DMEM with normal glucose, glutamine, and pyruvate (Life Technologies) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, gentamicin sulfate (50 g/ml), and amphotericin B (5 g/ml) at 37°C in 10% CO 2 . The medium was changed every 3 days to remove dead and non-attached cells until fibroblasts reached confluence. Monolayer cultures were maintained in the same medium. Lung fibroblasts were used between the second and fourth passages in all experiments. The purity of isolated lung fibroblasts was determined by crystal violet staining and by immunofluorescence staining using monoclonal antibody 3C4 against human fibroblasts as described previously (137). All of the treatments and transfection experiments were done with cells that were serum-starved for 24 h.
For SDS-PAGE, the denatured cell lysates were loaded onto a 4 -12% gradient polyacrylamide gel at 15-30 g of protein/lane in an Invitrogen minigel apparatus. Proteins were transferred to nitrocellulose membranes and blocked for 1 h with 5% nonfat dry milk in Tris-buffered saline containing 0.1% Tween 20 followed by washing in the same Tris-Tween buffer. The membranes were probed with the appropriate antibody diluted in Tris-buffered saline containing 5% bovine serum albumin (for polyclonal antibodies) or 5% nonfat dry milk (for monoclonal antibodies) followed by treatment with peroxidase-linked secondary antibodies and luminol reagents. The proteins on the blots were detected with antibodies for NOX4, NOX2, fibronectin, collagen-1, CD44V6, ␣-SMA phospho-Akt, and Akt (42,43,45,48,54,138). ␤-Tubulin and ␤-actin were used as internal standards. Sizes of proteins were estimated from prestained molecular weight standards electrophoresed in the same gel as the samples. Immunoreactive bands were quantified by densitometry. Each protein sample was analyzed from at least three independent experiments from each set of fibroblasts.

Immunohistochemical staining in lung sections
Lung sections from bleomycin-and saline-treated mice and from IPF patients and normal subjects were deparaffinized using standard procedures and permeabilized with 0.1% Triton X-100 in PBS. NOX4 and CD44V6 were localized in sections by immunohistochemical staining using NOX4 and CD44V6 antibodies from Abcam following the manufacturer's protocols. As a negative control, the primary antibody was replaced with nonimmune rabbit IgG, and no staining occurred.

CD44V6 shRNA and V6-PEP blocking peptide nanoparticle delivery in the bleomycin injury lung model of mice
The sequences of the two V6 human peptides are KEQWF-GNRWHEGYR (V6-PEP) and QATPSSTTEETATQ (V6-PEP-1). The sequence of the v6-murine (rat) peptide is KEKWFE-NEWQGKNP (v6-PEP-2). These peptides and the control peptide were prepared by Dharmacon Inc. The pSicoR-V6 shRNA/nanoparticles, the control non-targeting pSicoR-Scrambled shRNA/nanoparticles (control shRNA/nanoparticle), the V6-PEP/nanoparticles, and the control peptide/ nanoparticles were prepared as described in our published paper (45). For in vivo studies, the peptide/nanoparticle and shRNA/nanoparticle were used at doses of ϳ100 g/mouse by intratracheal delivery with bleomycin in a total volume of 50 l. The dose for shRNA/nanoparticles and for peptide/nanoparticles (ϳ100 g/mouse) was determined from a pilot dose-response experiment (data not shown and as described in our previous study (45)).

RNA silencing
The shRNA-coding nucleotide sequences of the genes were obtained from the NCBI, National Institutes of Health, website (www.ncbi.nlm.nih.gov), 4 and we designed hairpin shRNAs to target the transcript sequences using the Broad Institute GPP Web Portal (http://portals.broadinstitute.org/gpp/public/). 4 Sequences for cloning in pSico/pSicoR vectors were designed following the MIT Jacks Laboratory website (http://web.mit. edu/jacks-lab/protocols/pSico.html). 4 For example, the sense sequence for CD44v6 for cloning in pSico/pSicoR vector was 5Ј-TTAGTAGTACAACGGAAGAAACTTCAAGAGAGTTT-CTTCCGTTGTACTACTA-3Ј.
Similarly, the scrambled shRNA sequence obtained from the website was cloned in pSico/pSicoR vector. c-Fos shRNA and c-Fun shRNA were purchased from Dharmacon, as were the appropriate scrambled controls. Double-stranded oligonucleotide cassettes for control shRNA, CD44V6 shRNA, and NOX4 shRNA were prepared. The linearized pSicoR vectors were ligated to the double-stranded oligonucleotide cassettes (45). The resulting pSicoR-CD44v6 shRNA (v6 shRNA) and pSicoR-NOX4 shRNA transfectants constitutively silence the respective CD44V6 and NOX4 genes in the cells. pSicoR-scrambled shRNA (control shRNA) transfectants were used as control to the above shRNA transfectants. In the in vivo experiments, we used these shRNAs in Tf-PEG-PEI (nanoparticles) following our published paper (45).

Confirming the specificity of shRNA experiments
To confirm the shRNA knockdown efficiencies, in specific experiments, we used more than one shRNA. We also confirmed the knockdown experiments, comparing the knockdown effects of shRNAs for CDS either with those of NCDS (as proper negative controls) or with rescue of the observed shRNA-mediated knockdown phenotype by expression of a resistant form of the targeted mRNA. This is performed by 1) transfecting the cells with specific shRNAs for the CDS of the target gene or 2) co-transfecting the shRNA (CDS) for the target gene with or without corresponding cDNA transfection or by the indicated shRNA-mediated knockdown and corresponding KI gene transfection (see Table 1 and Fig. 12 (A-F) in our companion paper) (71). Total cell lysates were examined by Western blot analysis for the indicated proteins and ␤-tubulin. Total mRNA were analyzed for the indicated mRNAs by RT-PCR and real-time PCR. Synthetic shRNAs that are 21-23 nucleotides in length have been shown to effectively silence specific target genes by promoting mRNA degradation in cultured mammalian cells and mice (45,139,140). However, in cultured cells, one potential source of off-target effects by either transfected shRNA duplexes or endogenously expressed shRNAs is the unintentional activation of the interferon response (141)(142)(143). Previous work has shown that nonspecific gene target effects that include off-target gene suppression and up-regulation of type I IFN response are induced by dsRNAs Ն 30 bp in length (144). There are also recent reports of shRNA-induced stimulation of Toll-like receptors (TLR3 (dsRNA) and TLR9 (unmethylated CpG)) and downstream IFN␣ and IFN␤ responses (145). Despite the highly publicized in vivo miRNA-related toxicity of shRNA, it has been shown that shRNA has several advantages over siRNA: fewer off-target effects, multiple-target-silencing capacity without a corresponding increase in dose, durability of effect, and inducible application (146,147). Because we use siRNAs and shRNAs extensively to study CD44V6-induced fibrogenic functions, we follow several steps (as shown in our companion paper (71) and in Fig. 2 (A-F)) that can avoid this off-target problems as well as confirm the specificity of the shRNAs used in this study. Specifically, the confirmation of knockdown by siRNA and shRNAs was done by using more than one shRNA by demonstrating that the expression level of the target mRNA was substantially reduced, whereas the level of expression of a control shRNA (scrambled shRNA)-transfected mRNA and/or protein was unaffected (Figs. 2 (A, B, E, and F), 4 (A and B), 5 (B-G), 6 (B-D), and 7 (D-G)) and in our companion paper (71).

In vivo mouse studies for measuring IFN
C57BL/6 mice (6 -8 weeks of age; Jackson Laboratories) were used for the shRNA injections. The shRNAs were formulated in D5W (5% (w/v) glucose in water). To examine the effect of various types of administration, mice received 2.5 mg/kg shRNA either i.p. or through the tail vein via LPTV (1% (v/w)) or HPTV (10% (v/w)). The HPTV method of administrating shRNA provides for cellular uptake and gene down-regulation in the livers of mice (140,148). Mouse tails were warmed with a heating pad before injection of a 0.2-ml (for LPTV) or a 2-ml (for HPTV) volume (per 20 g of mouse weight) over ϳ5 s. To measure plasma cytokine levels, blood was harvested from mice 2 h after injection by cardiac puncture, and plasma was isolated using Microtainer tubes (BD Biosciences). Preliminary experiments compared the IFN-␣ response with poly(I:C) at 1, 3, 6, 12, and 24 h after injection, and the maximum response was consistently observed at 3 h. Therefore, data from the 3-h time point after shRNA treatment were used. IFN-␣ levels were measured by ELISA according to the manufacturer's instructions.

Transient transfection using HNLFbs and IPFFbs
All transfections were done using Lipofectamine (Invitrogen) in cultures at ϳ75% confluence. After transfection, the cultures were grown for another 72-96 h for analyses.

mRNA expression analysis by RT-PCR
Preparation of mRNA from tissue samples used a commercially available mRNA purification kit. cDNA was synthesized with mRNA as template for an oligo(dT)-primed reaction catalyzed by reverse transcriptase. The quality of mRNA preparations and cDNA syntheses was checked by including GAPDHor ␤-actin-specific primers as an internal RT-PCR control. All cDNA sequences were obtained from the GenBank TM database. RT-PCR used human gene-specific sense and antisense primers based on sequences published in GenBank TM . PCR was done for 30 cycles for GAPDH, 32 cycles for p22 phox , and 36 cycles for all other primers, with amplification at 95°C for 1 min, 58°C for 1 min, and 72°C for 2 min, followed by an extension step at 72°C for 10 min. For semiquantitative RT-PCR, PCR-amplified DNA was separated on 1.2% agarose gels, stained with ethidium bromide, and visualized and photographed under UV light.

Real-time PCR of CD44v6 in MNLFbs and HNLFbs
Total RNA was isolated from MNLFbs and HNLFbs after various treatments, and transfections were done as mentioned in the figure legends for each specified experiment using the RNeasy minikit (Qiagen) according to the standard protocol provided by the manufacturer, with on-column DNA digestion. RNA integrity and concentration were analyzed using Bioanalyzer (City State), and 1 g of RNA was reverse-transcribed into cDNA using the First Strand cDNA synthesis kit from Roche Applied Science. SYBR Green (Bio-Rad) was used for all realtime PCR assays. Amplification was done with the real-time PCR analyzer (Bio-Rad). The PCR mixture (25 l) contained 12.5 l of 2ϫ SYBR Green PCR Master Mix (Bio-Rad), 5 l of diluted RT product (1:20), and 0.5 M sense and antisense primer sets. The primers used for human and mouse genes are shown in Table 1. The real-time PCR assays were done in three individual experiments with triplicate samples using standard conditions. After sequential incubations at 50°C for 2 min and 95°C for 10 min, respectively, the amplification protocol consisted of 50 cycles of denaturing at 95°C for 15 s, annealing, and extension at 60°C for 60 s. The standard curve was made from a series dilution of template cDNA. Expression levels of ␣-SMA, NOX4, collagen-1 (COL1A1), and CD44V6 mRNAs were calculated after normalization with the housekeeping gene GAPDH.

AP-1 transcription factor-binding activity
Nuclear extracts were prepared with the NucBuster protein extraction kit (Novagen) according to the manufacturer's protocol. AP-1 DNA-binding activity was measured using the TransBinding AP-1 assay kit (Panomics). Briefly, nuclear extracts were incubated with biotinylated AP-1 consensusbinding sequence oligonucleotides, and the complexes bound to the oligonucleotides were detected using a primary AP-1 antibody and a secondary HRP-conjugated antibody. AP-1 activity was assayed by measuring the absorbance at 450 nm.

Transient transfection
All transfections were done using Lipofectamine (Invitrogen) in cultures that were ϳ75% confluent. After transfection, the cells were cultured for 72-96 h prior to analyses.

Cell proliferation assays
Cell proliferation was measured by a BrdU cellular ELISA kit based on the incorporation of the pyrimidine analogue BrdU into the DNA of proliferating cells that are cultured in microtiter plates. After its incorporation into DNA, BrdU in the cells was detected by anti-BrdU monoclonal antibody. Fibroblasts (20,000) were cultured in 96-cell plates at 37°C. BrdU (20 M) was added to wells during the final 2-18 h of culture. Cells were fixed and permeabilized, and the DNA was denatured to enable antibody binding to the incorporated BrdU. Detector mouse anti-BrdU monoclonal antibody was pipetted into the wells and allowed to incubate for 1 h. unbound antibody was removed, HRP-conjugated goat anti-mouse IgG was added, and then tetramethylbenzidine was added. The absorbance was detected at 450 nm and was quantified in a microplate reader (BioTek, Winooski, VT).

Migration of leukocytes through endothelial cell cultures
Leukocytes patrol the vascular system, and their rolling, firm attachment, and subsequent migration across endothelial walls are necessary for their recruitment to sites of inflammation. This process involves a multistep cascade consisting of leukocyte rolling, adhesion, and transmigration. A quantitative assay for leukocyte transendothelial migration has been described using a modified Boyden chamber system (149). This is a twochamber system with a porous membrane providing an interface between the two chambers. Mouse endothelial cells (C166 from ATCC) were cultured on top of the porous membrane that was coated with an ECM protein. Isolation of PBMCs (leukocytes) from mice was done by density gradient centrifugation of buffy coats using Lympholyte-M (Atlanta Biologicals) gradient medium, according to the manufacturer's instructions. Isolated lung fibroblasts from mice on days 0 (MNLFbs) and 21 after bleomycin treatment (21dBLMFbs) were used in these experiments as described (see Fig. 1 in our companion paper (71)). Migration of the leukocytes across the endothelial monolayer was determined by adding confluent leukocytes 10 ϫ 10 3 cells/6.5-mm-diameter transwell culture insert (8-m pore size) on the top of a confluent monolayer (10 ϫ 10 3 endothelial cells). The effect of fibroblasts in the bottom chamber on leukocyte migration was examined by co-culturing lung fibroblast monolayers (MNLFbs and 21dBLMFbs; 10 ϫ 10 3 ) in the tissue culture wells below the transwell insert. Leukocyte migration across the endothelial cell layer in the upper chamber was determined by measuring the number of cells that migrate between the endothelial cells, through the porous membrane, and into the lower chamber. The cell numbers present on the undersurface of the transwell insert were determined by direct counting under a dissecting microscope. Medium from the lower chamber with the loosely adherent leukocytes from the floor of the lower chamber was collected with fresh DMEM. The resulting suspension of cells was centrifuged at 3000 ϫ g for 5 min and resuspended in counting fluid. Cell numbers were determined by direct counting using a hemocytometer. Based on the results of the initial experiments, a 16-h migration period was found to give consistent results, and this time point was selected for subsequent studies.

Cell adhesion assay
Cells from the cell layer that adhered onto the inside part of the insert were isolated, and 100 l of the cell suspension/well was added onto 96-well plates in DMEM with 0.1% fatty acidfree bovine serum albumin followed by incubation for 30 min at 37°C under 5% CO 2 with the lid off. After incubation, unattached cells were removed by rinsing with PBS. Attached cells were fixed in 5% glutaraldehyde for 20 min and stained with 0.1% crystal violet. After washing, stains were dissolved in 200 l of 10% acetic acid, and color was read at 575 nm in a BioTek plate reader.

Invasion assay
As described in detail previously (49,54,150), the invasive potential of the transfected cells was evaluated by measuring the number of cells invading the Matrigel-coated transwell chambers (BD Biosciences).

Cytotoxicity assay
2 ϫ 10 5 PBMCs were treated with the synthetic peptide/ nanoparticles or the shRNA/nanoparticles at 1, 10, and 100 M for 24 h. Then 10 l of MTT (Calbiochem) solution (5 mg/ml) was added to each well, and the culture was further incubated for 4 h at 37°C. A total of 50 l of N-dimethylmethanamide solution was added to each well, followed by overnight incubation. The absorbance of each well was determined on a microplate reader at 592 nm. The relative cell viability (percentage) was calculated using the formula, % viability ϭ (A 592 (treated cells)/(A 592 (untreated cells)) ϫ 100. Negative control cells were treated with medium.

Preparation of conditioned media
Monolayers of fibroblasts in 75-cm 2 tissue culture flasks were rinsed twice with serum-supplemented (10% FBS) CD44V6, NOX4, and TGF␤1 signaling in IPF DMEM, and 10 ml of fresh medium was added. After a 24-h conditioning period, media were harvested, and cellular debris was removed by centrifugation at 3000 ϫ g for 10 min. Media were either used immediately or stored frozen at Ϫ80°C for later experiments.

V6-PEP detection by ELISA
2 ϫ 10 6 PBMCs diluted (1:1 and 1:2 dilution) in bicarbonate buffer (0.1 M, pH 9.4) were coated in a 96-well Maxisorp microtiter plate (Nunc, ThermoFisher Scientific) and incubated overnight at 4°C. After blocking with 5% BSA in PBS (Sigma-Aldrich) at 37°C for 1 h, each well was washed once with PBS. V6-PEP (1 g/ml) or TGF␤1 (1 ng/ml) was added, and the plate was incubated at 37°C for 1 h. Each well was washed five times with PBS, and anti-TGF␤1 antibody labeled with HRP (BD Bioscience) diluted at 1:250 was added into each well, followed by incubation for 1 h at 4°C. Each well was washed five times and treated with o-phenylenediamine dihydrochloride 1 mg/ml (Sigma-Aldrich), a soluble substrate for the detection of peroxidase activity in ELISA. The reaction was stopped by adding 20 l of 4 N H 2 SO 4 and read at 492 nm in a microplate reader.

Intracellular ROS detection
Fibroblasts grown in glass-bottom plates were washed with Hanks' buffered salt solution and loaded with H 2 DCFDA (Invitrogen) dissolved in Hanks' buffered saline at 10 M. Fluorescence was read at 535 nm using a plate reader.

Three-dimensional collagen gel contraction assay
HNLFbs were treated or transfected with various reagents as mentioned in the figure legends for the specified experiment. Collagen gels were reconstituted by mixing one part 3 mg/ml neutralized rat tail collagen-1 and two parts cell suspension in serum-free medium. Cell suspensions were seeded at a density of 200,000 cells/ml into 4-well tissue culture plates, and the gels were allowed to polymerize at 37°C for 1 h before adding 1 ml of medium. A volume of 600 l/gel was fabricated in a 4-well plastic culture dish, which ensured that the gel would remain attached throughout the culture period. The stressed HNLFbs were cultured in 10% FBS for 20 h after polymerization. The stressed HNLFbs were then incubated in 0.5% FBS with or without TGF␤1 (5 ng/ml) for another 48 h in serum-deprived medium, and the edges of the gels were then gently detached from the walls of the well using a sterile spatula. Gels were then photographed after 16 h, and the areas were measured using ImageJ software (National Institutes of Health).

In vivo targeting of CD44V6 by genetic modification and with the CD44V6-blocking V6-PEP
For genetic targeting of CD44V6, we planned to test whether intratracheal treatment with engineered nanoparticle delivery systems to convey CD44V6 shRNA in the fibrogenic lung cells. The idea was to transactivate a conditionally silenced plasmid with CD44V6 oligonucleotide by Cre recombinase produced in response to a tissue-specific promoter. The principle is as follows. The recombinase, produced under the influence of a tissue-specific promoter in the cells will eliminate the (CMV-EGFP)-cassette from U6-(CMV-EGFP) f/f -CD44V6 shRNA from the pSico-CD44V6 shRNA, and the U6 promoter will induce synthesis of CD44V6 shRNA. Normal cells in different organs will not be affected because they rely mostly on the standard CD44s expression, which does not have any variant exons. Even if the cells express CD44V6, shRNA will not be produced due lack of response to the tissue-specific promoter. The plasmids will be delivered through nanoparticle (Tf-PEG-PEI) carrier. Unused plasmids in other organs will be progressively destroyed by cytoplasmic nucleases, and the nanoparticles will be cleared because PEG and PEI are biodegradable, thus avoiding any toxicity problem. It has been shown that FSP-1 increases in the lung fibroblast cells after bleomycin treatment (151,152). A recent study also indicates that conditional deletion of Has2 by use of the FSP-1-Cre mouse line in mesenchymal lung fibroblast cells abrogated the invasive fibroblast phenotype, impeded myofibroblast accumulation, and inhibited the development of lung fibrosis (19). Thus, to assess the contribution of fibroblast expression of HAS2/CD44v6, we used a FSP-1 promoter-driven Cre plasmid. Further, using our nontoxic and non-inflammatory shRNA/nanoparticle delivery method (45,109), V6-PEP/nanoparticles, CD44V6 shRNA/ nanoparticles (pSico-CD44v6 shRNA/Tf-PEG-PEI plus FSP-1-Cre/Tf-PEG-PEI), control peptide/nanoparticles, or control shRNA/nanoparticle (pSico-scrambled shRNA/Tf-PEG-PEI plus FSP-1-Cre/Tf-PEG-PEI) were administered every other day from day 2 to day 30 by intratracheal delivery to the lungs of young mice during the onset of inflammation by bleomycin injury, and then the mice were left untreated for another 15 days. Lung tissue was harvested at the indicated times up to 45 days after injury. The time course of fibrosis induction, treatment schedule, and resolution of fibrosis was measured.

Preparation of pSico-CD44v6 shRNA and pFSP-1-Cre
Preparation of Tf-coated PEG-PEI nanoparticles and nanoparticle size determination were carried out as validated by our group previously in an intestinal/colon cancer murine model (45,109,153).

Sircol assay for collagen
Acid-soluble collagen in whole-lung homogenates was analyzed by the Sircol assay as described previously (154).

Hydroxyproline content of whole lung
A lobe of each mouse lung was homogenized in PBS and then acidified by adding an equal volume of 12 N HCl, hydrolyzed by heating at 120°C for 24 h, and processed for hydroxyproline measurements as described previously (155).

Lung histology staining
Paraffin-embedded tissue sections were processed for Masson's trichrome staining for antibodies against CD44V6 and NOX4 as described previously (1).

Measurement of hydrogen peroxide production
Extracellular H 2 O 2 released from cultured cells was assayed by Amplex Red. The Amplex Red H 2 O 2 assay kit uses the Amplex Red reagent (10-acetyl-3,7-dihydroxyphenoxazine) to detect H 2 O 2 released from cells. In the presence of HRP, the Amplex Red reagent reacts with H 2 O 2 in a 1:1 stoichiometry to produce the red-fluorescent oxidation product, resorufin. Fluorescence was read at 585 nm using a plate reader.

Proximity ligation Duolink assay
We used the Duolink in situ PLA reagent from Olink Biosciences (Uppsala, Sweden) to characterize interaction between endogenous CD44v6 and TGF␤1RI. Cells were grown on glassbottom culture dishes (MatTek) according to the protocol supplied by Duolink reagents from Sigma-Aldrich. After incubating overnight with the CD44V6 and TGF␤RI antibodies, the cells were further incubated with the oligonucleotide labeled with anti-goat plus and anti-mouse minus PLA probes, as recommended by the manufacturer. Negative control slides were incubated with secondary antibodies only before incubation with PLA probes. Samples were mounted with the Duolink mounting medium containing DAPI to counterstain nuclear DNA. PLA images (fluorescence dots) were acquired using a Zeiss LSM 880 NLO confocal microscope (Cell and Molecular Imaging Shared Resource, Hollings Cancer Center, Medical University of South Carolina) with a 63ϫ/1.4 numeric aperture objective and analyzed by the Duolink Image Tool (Sigma). Counted dots are expressed as mean Ϯ S.E.

Statistical analysis
Each experiment was done in triplicate and was repeated three times (n ϭ 9). Data were pooled for statistical analyses. Animal studies, Western blot analyses, mRNA analyses, and proliferation experiments for each separate experiment were repeated three or four times. Data from various groups are expressed as means Ϯ S.E. (n ϭ 3-4). Statistical comparisons were done using Student's t test for unpaired samples or using two-way analysis of variance (ANOVA) with Bonferroni posttest correction on grouped data as mentioned in the figure legends.