Wnt Coreceptor Lrp5 Is a Driver of Idiopathic Pulmonary Fibrosis

Abstract

Rationale: Wnt/β-catenin signaling has been implicated in lung fibrosis, but how this occurs and whether expression changes in Wnt pathway components predict disease progression is unknown.

Objectives: To determine whether the Wnt coreceptor Lrp5 drives pulmonary fibrosis in mice and is predictive of disease severity in humans.

Methods: We examined mice with impaired Wnt signaling caused by loss of the Wnt coreceptor Lrp5 in models of lung fibrosis induced by bleomycin or an adenovirus encoding an active form of transforming growth factor (TGF)-β. We also analyzed gene expression in peripheral blood mononuclear cells (PBMC) from patients with idiopathic pulmonary fibrosis (IPF).

Measurements and Main Results: In patients with IPF, analysis of peripheral blood mononuclear cells revealed that elevation of positive regulators, Lrp5 and 6, was independently associated with disease progression. LRP5 was also associated with disease severity at presentation in an additional cohort of patients with IPF. Lrp5 null mice were protected against bleomycin-induced pulmonary fibrosis, an effect that was phenocopied by direct inhibition of β-catenin signaling by the small molecular inhibitor of β-catenin responsive transcription. Transplantation of Lrp5 null bone marrow cells into wild-type mice did not limit fibrosis. Instead, Lrp5 loss was associated with reduced TGF-β production by alveolar type 2 cells and leukocytes. Consistent with a role of Lrp5 in the activation of TGF-β, Lrp5 null mice were not protected against lung fibrosis induced by TGF-β.

Conclusions: We show that the Wnt coreceptor, Lrp5, is a genetic driver of lung fibrosis in mice and a marker of disease progression and severity in humans with IPF. Evidence that TGF-β signaling can override a loss in Lrp5 has implications for patient selection and timing of Wnt pathway inhibitors in lung fibrosis.

At a Glance Commentary

Scientific Knowledge on the Subject

Idiopathic pulmonary fibrosis (IPF) is a chronic interstitial lung disease that affects 5 million people worldwide and is associated with high mortality rates. Other than lung transplantation, no therapies have shown a survival benefit in IPF, in part because of a gap in understanding of the molecular pathways driving this disease.

What This Study Adds to the Field

This study shows that expression analysis of peripheral blood mononuclear cells from patients with IPF reveals Wnt signaling as one of the top two overrepresented pathways associated with disease progression. We also show that loss of the Wnt coreceptor, Lrp5, is protective in the bleomycin but not transforming growth factor-β model of murine fibrosis, suggesting that Wnt pathway inhibitor strategies should be considered judiciously with regards to the degree of transforming growth factor-β activation.

Idiopathic pulmonary fibrosis (IPF) is a progressive disease with mortality rates that exceed those of many malignancies. Although familial forms of lung fibrosis are caused by mutations in proteins that are seemingly unrelated (e.g., surfactant proteins A2 and C, and telomerase, reviewed in [1]), gene expression profiling of IPF lungs reveals striking similarities to the developing lung, suggesting that aberrant activation of developmental signaling pathways may contribute to the persistent nature of IPF (2). Indeed, a prevailing model in the field reasons that epithelial injuries trigger fibroblast-dependent repair processes that patients with IPF cannot resolve normally, which leads to excessive fibroproliferation and matrix deposition at the expense of normal tissue remodeling. Although dysregulated developmental signaling pathways may not be the actual trigger of fibrosis, there is great interest in assessing their contribution to fibrosis, given the availability of small molecule therapeutics that target these pathways.

The Wnt/β-catenin signaling pathway is one such pathway found dysregulated in microarrays from patients with lung fibrosis (2). Wnt proteins comprise a family of secreted molecules that instruct cells to adopt particular fates throughout development and adult tissue homeostasis. Cellular identity is ultimately controlled by a transcription complex that contains a DNA-binding factor known as lymphocyte enhancer factor/T-cell factor, and the dual signaling/adhesion protein, β-catenin (3). Although Wnt/β-catenin signaling occurs in all tissues, the genes activated by this transcriptional complex are cell-type and -context dependent (4), and the lung-specific targets of this pathway are largely undefined (5). Previous studies have shown evidence of Wnt/β-catenin signaling up-regulation in lung samples from patients with IPF (6, 7). Although pharmacologic and genetic inhibitors of the Wnt/β-catenin pathway (8–10) or inhibition of one target of β-catenin signaling (11) can limit the development of fibrosis in mice, the cellular and molecular mechanisms by which Wnt/β-catenin signaling drives fibrosis are not known. Moreover, the extent to which Wnt/β-catenin signaling predicts outcome in patients with IPF has remained unexamined. To address these questions, we analyzed peripheral blood mononuclear cells (PBMC) from patients with IPF, and undertook a series of experiments using a viable mouse with reduced Wnt/β-catenin signaling resulting from loss of one of its obligate coreceptors, LRP5, in two murine models of lung fibrosis. Some of the results of these studies have been previously reported in the form of abstracts (12, 13).

Methods

Design of Human Study

The PBMC microarray cohort consisted of 74 patients recruited and prospectively followed at the University of Pittsburgh. The findings were validated in a separate cohort of 45 patients evaluated at the University of Chicago. IPF diagnosis was established by a multidisciplinary group using American Thoracic Society/European Respiratory Society criteria (14) and was in accordance with the most recent guidelines (15). Patients were excluded from the study if they had evidence of autoimmune syndromes, malignancies, infections, drugs, or occupational exposures known to cause lung fibrosis. The studies were approved by the Institutional Review Boards and informed consent was obtained from all patients at each institution. Demographic and clinical information, including spirometry and diffusion capacity of the lung for carbon monoxide (Dl_CO), was collected at baseline and every 3 months (on average). The time-to-event outcome analyzed was disease progression, defined as either a 10% relative decline in FVC (16, 17), 15% relative decline in Dl_CO (17), or a 5% absolute increase in the patient Composite Physiologic Index (CPI) (17), whichever happened first. In this analysis, the first observance of any of these events was counted as an outcome. Three patients in the Pittsburgh cohort were censored at their first visit because they did not have follow-up spirometric and Dl_CO data available. To validate our findings, and given the lack of longitudinal pulmonary function data in other publically available PBMC microarray cohorts, we correlated the expression of WNT pathway genes with spirometric and Dl_CO markers at presentation in patients evaluated at the University of Chicago. The gene expression data were included in a previous manuscript (18) and are publically available in the gene expression Omnibus as GSE28221.

Methodology for the bleomycin model of murine fibrosis, collagen quantification, primers for quantitative polymerase chain reaction (qPCR), and cell culture studies can be found in the online supplement.

Results

Expression of Wnt Pathway Components in PBMC Is Associated with IPF Progression

To ask whether expression changes in Wnt signaling components predict IPF disease progression, we reanalyzed a recently published microarray cohort (18) from PBMC of 74 patients with IPF evaluated at the University of Pittsburgh using gene-set analysis with censored outcome data. We defined disease progression as either a 10% relative decline in FVC (16, 17), 15% relative decline in Dl_CO (17), or a 5% absolute increase in the patient CPI (17), whichever happened first, during patient follow-up. Demographic and clinical characteristics of the IPF microarray cohort are detailed in Table 1. Gene-set analysis identified 13 pathways (false discovery rate < 5% Maxmean score ≥ 0.5 and ≤−0.5) associated with disease progression (see Table E1 in the online supplement). Among them, the “WNT signaling” Pathway Interaction Database gene set was one of the top two overrepresented pathways associated with disease progression (Maxmean score of 0.834) (19). LRP5 and LRP6 were among the genes of this pathway with the strongest association with disease progression when overexpressed (Cox scores, 0.77 and 1.095, respectively) (Figure 1A). We further explored the association between microarray expression of LRP5 and LRP6 with disease progression using Kaplan-Meier curves. We identified that LRP5 and LRP6 overexpression was associated with IPF progression in the subset of patients with higher expression levels of each gene (split at the 84th and 80th percentile expression level, respectively) (Figures 1B and 1C). Hazard ratios were similar for subjects with high LRP5 (hazard ratio, 2.38; P = 0.0051; 95% confidence interval [CI], 1.00–5.64) and high LRP6 (hazard ratio, 2.24; P = 0.0086; 95% CI, 0.96–5.2). The median time to disease progression in the subset of patients with higher LRP5 expression was 3 months versus 5.7 months in patients with lower expression levels. Similarly, patients with high LRP6 expression had median times to disease progression that were shorter (3.45 mo) than subjects with low LRP6 expression (5.68 mo). LRP5 and LRP6 were also significantly (P < 0.05) associated with disease progression after adjusting their expression levels for demographic variables, such as age, sex, smoking history, and use of immunosuppressive medications (see Tables E2–E5). We then studied the correlation between LRP5 and LRP6 gene expression levels at presentation with clinical markers of disease severity in an additional PBMC microarray cohort of patients with IPF evaluated at the University of Chicago. We identified that LRP5 expression levels at presentation were significantly correlated with Dl_CO% (r = −0.43; 95% CI, −0.63 to −0; P = 0.001) and CPI (r = 0.44; 95% CI, 0.17–0.65; P = 0.002) in patients with IPF in this cohort (Figures 1D and 1E). Although 2 out of 45 patients from the University of Chicago cohort were on immunosuppressive drugs at the time of blood draw, LRP5 is still significantly correlated with Dl_CO% and CPI (P < 0.05) after excluding these two patients from the correlation analysis. LRP6 expression levels were not significantly correlated (P > 0.05) with spirometric measures, Dl_CO%, or CPI in the University of Chicago cohort.

Table 1.

Demographic and Clinical Characteristics of the Idiopathic Pulmonary Fibrosis Microarray Cohorts

Characteristic	University of Pittsburgh Cohort (n = 74)	University of Chicago Cohort (n = 45)
Age, yr
Mean ± SD	69.1 ± 8.1	66.9 ± 8.1
Gender, n (%)
Males	51 (68.9)	40 (88.9)
Females	23 (31.1)	5 (11.1)
Race, n (%)
White	72 (97.4)	37 (82.2)
Black	1 (1.3)	3 (6.7)
Asian	1 (1.3)	0 (0)
Hispanic	0 (0)	5 (11.1)
Smoking status, n (%)
Ever smoker	46 (62.2)	27 (60)
Never smoker	28 (37.8)	18 (40)
Pulmonary function tests (mean ± SD)
FVC, %	65 ± 16	62 ± 14
DlCO, %	49 ± 18	44 ± 17
FEV1, %	76 ± 17	75 ± 16
CPI	51 ± 14	56 ± 13

Definition of abbreviations: CPI = Composite Physiologic Index; Dl_CO% = diffusing capacity of carbon monoxide percent predicted.

Figure 1.

Wnt signaling pathway components are associated with the progression of idiopathic pulmonary fibrosis (IPF). (A) Peripheral blood mononuclear cells from 74 subjects with IPF in the University of Pittsburgh cohort were analyzed for expression of Wnt pathway genes. A positive Cox score indicates that higher expression correlates with rapid IPF progression, whereas lower expression indicates slower IPF progression. A negative Cox score indicates that higher expression correlates with slower IPF progression, whereas lower expression correlates with rapid IPF progression. Genes labeled in red are consistent activators of β-catenin signaling; genes labeled in green are consistent negative regulators of β-catenin signaling. *Wnt pathway components with contextual effects. (B and C) Kaplan-Meier plots for LRP5 (B) and LRP6 (C). Red dotted lines represent patients with microarray expression levels above the 84th percentile for LRP5 and above the 80th percentile for LRP6. Black lines represent patients with expression levels below the threshold of these two genes. The median time to disease progression for each group is depicted in dotted vertical lines. (D and E) Correlation between the peripheral blood mononuclear cell microarray expression of LRP5 at presentation, diffusion capacity of the lung for carbon monoxide (Dl_CO) % (D) and Composite Physiologic Index (CPI) (E) in patients with IPF evaluated at the University of Chicago. Statistics are provided in the text.

Mice Lacking Lrp5 Are Protected from Bleomycin-induced Lung Fibrosis

To understand how Wnt signaling contributes to lung fibrosis, we examined mice with impaired Wnt signaling caused by loss of the Wnt coreceptor Lrp5 (20) in the bleomycin model of lung fibrosis. Although Lrp5 is expressed in the normal adult mouse lung, no gross abnormalities are observed in Lrp5^−/− lungs (Figure 2A), which is likely the result of continued signaling through its paralog Lrp6 (21). Twenty-one days after the intratracheal administration of bleomycin, mice lacking Lrp5 show a marked attenuation of tissue injury and matrix deposition compared with wild-type Lrp5^+/+ littermates by histology using Masson trichrome and hematoxylin and eosin stains (Figure 2A; see Figure E1). Quantification of total protein within lung airspaces using the bronchoalveolar lavage (BAL) method supports this qualitative assessment of lung injury, because BAL fluid from Lrp5^−/− mice contains less protein (1,957 ± 562 μg/ml; 691 ± 171 μg/ml) than their Lrp5^+/+ counterparts (Figure 2B). Quantification of total soluble collagens from Lrp5^+/+ and null lungs using the Pico-Sirius red dye method also supports the Trichrome in situ staining of matrix collagen, because soluble collagen content in Lrp5^+/+ lungs increases after bleomycin treatment (75.5 ± 14.9 μg/ml, 133.6 ± 31.9 μg/ml) but this increase is significantly blunted (75.6 ± 16.1 μg/ml, 86.9 ± 23.2 μg/ml) in Lrp5^−/− mice (Figure 2C). This finding is confirmed at the level of newly synthesized type 1 collagen mRNA, where Lrp5 loss markedly diminishes bleomycin-mediated up-regulation of Col1A1 (18.1 ± 0.12-fold, 5.9 ± 0.19-fold) and Col1A2 (7.4-fold, 2.2-fold) genes by qPCR (Figure 2D). This reduction in collagen content measured 21 days postinjury does not correlate with differences in peak lung injury assessed 6 days postinjury (see Figure E2). In addition, no significant differences in α-smooth muscle actin were quantified by qPCR or immunohistochemistry 21 days post bleomycin injury, as often described for this model (22).

Figure 2.

Lrp5 loss attenuates bleomycin-induced pulmonary fibrosis. (A) Masson trichrome staining of Lrp5 wild-type (Lrp5^+/+) and null (Lrp5^−/−) lungs instilled with bleomycin or saline (control) and processed for histology 21 days postinjury. Boxed region is shown at higher magnification. Bar = 400 μm. Corresponding hematoxylin and eosin stained lungs are shown in Figure E1. (B) Total protein in bronchoalveolar lavage fluid from Lrp5^+/+ and Lrp5^−/− mice treated with saline (black bars) or bleomycin (patterned bars) 21 days postinjury. n = 6 animals/condition. (C) Total soluble lung collagen by Sircol method harvested from mice 21 days postinjury. Number greater than or equal to four animals per condition. (D) Col1A1 and Col1A2 mRNA expression by quantitative polymerase chain reaction and normalized to glyceraldehyde phosphate dehydrogenase (GAPDH) from lungs harvested 21 days postinjury. Results are expressed as the mean gene expression ± SD relative to GAPDH from greater than or equal to four animals per condition and normalized to WT saline treatment. Relative change in gene expression was calculated using the 2^−ΔΔCt method. *P < 0.05, **P < 0.01, ***P < 0.001.

Direct attenuation of Wnt signaling using a small molecule inhibitor of β-catenin responsive transcription (iCRT3) (23) administered after peak injury (24) also attenuates lung injury and collagen expression (Figure 3). BAL fluid total protein decreased with iCRT3 treatment (1,357 ± 440 μg/ml, 793 ± 711 μg/ml), and soluble collagen content (1,957 ± 311 μg/ml, 1,574 ± 360 μg/ml) and Col1A1 and Col1A2 genes decreased to 0.4- and 0.5-fold, respectively. Primary lung fibroblasts derived from mice lacking Lrp5 are indeed less responsive to Wnt3a than Lrp5^+/+ cells (118-fold, 4.4-fold), but not to a constitutively active form of β-catenin (5.2-fold, 4.7-fold), demonstrating that Lrp5 loss indeed reduces sensitivity to Wnt ligands (Figure 4). Altogether, these data show that the Wnt coreceptor, Lrp5, is required for the development of fibrosis, likely through a reduction in Wnt/β-catenin signaling.

Figure 3.

Small compound inhibitor of β-catenin signaling, inhibitor of β-catenin responsive transcription (iCRT3), limits bleomycin-induced injury and collagen expression. (A) Hematoxylin and eosin staining of C57BL/6 mice receiving daily intraperitoneal iCRT3/Dulbecco’s modified Eagle medium (DMSO) or DMSO (vehicle control) alone starting 6 days after bleomycin administration via intratracheal route; lungs were harvested 21 days later and processed for histology. Boxed region is shown at higher magnification. Bar = 400 μm. (B) Total protein in bronchoalveolar lavage (BAL) fluid from iCRT3-treated and untreated mice subjected to saline (solid icons) or bleomycin (open icons) injury. Injections were daily (QD) or every other day (QOD). Each symbol represents data from a single mouse lung. *P < 0.05 by unpaired, two-tailed t test comparison of columns indicated with brackets. (C) Total soluble lung collagen by Sircol method harvested from mice 21 days postinjury and iCRT3 treatment. *P < 0.05 and ***P < 0.001 by analysis of variance comparison with saline + DMSO (control) treatment. (D) Col1A1 and Col1A2 mRNA expression by quantitative polymerase chain reaction and normalized to glyceraldehyde phosphate dehydrogenase (GAPDH) from lungs harvested 21 days postinjury. Results are expressed as the mean gene expression ± SD relative to GAPDH from greater than or equal to four animals per condition and the drug-treated condition is normalized to its corresponding saline or bleomycin control. Relative change in gene expression was calculated using the2^−ΔΔCt method. *P < 0.05 by unpaired t test of columns indicated with brackets.

Figure 4.

β-Catenin signaling is diminished in Lrp5 null fibroblasts. (A) Immunoblot of primary lung fibroblasts isolated from Lrp5^+/+ or Lrp5^−/− mice infected with Adeno-GFP, Adeno-Wnt3a, or Adeno-S37A–β-catenin for 36 hours. Activation of Wnt/β-catenin signaling was assessed by quantification of the cadherin-free pool of β-catenin using a GST-ICAT affinity matrix (59). Total inputs of the active form of β-catenin (ABC, which is largely cadherin-associated [60]) and glyceraldehyde phosphate dehydrogenase (GAPDH) are shown. (B and C) Activation of Wnt/β-catenin signaling assessed by quantification of the canonical target gene, AXIN2. Axin2 mRNA expression was normalized to GAPDH from Lrp5^+/+ or Lrp5^−/− lung fibroblasts infected with Adeno-GFP or Adeno-Wnt3a (B), and Adeno-GFP or Adeno-S37A-β-catenin (form of β-catenin refractory to GSK3-mediated phosphorylation and degradation [61]). Results are expressed as the mean gene expression ± SD relative to GAPDH from three independent infections of two separate fibroblast isolations and normalized to Lrp5^+/+ Adeno-GFP treatment. Relative change in gene expression was calculated using the2^−ΔΔCt method. ***P < 0.001 by analysis of variance comparison of Wnt3a activated Lrp5^+/+ versus Lrp5^−/− fibroblasts.

Lrp5 Loss in Bone Marrow–derived Cell Types Is Not Sufficient to Protect from Bleomycin-induced Fibrosis

To broadly define the cellular compartment where Lrp5 promotes fibrosis, we used a bone marrow–chimera approach. Wild-type mice transplanted with bone marrow cells from Lrp5^−/− or Lrp5^+/+ littermates demonstrate more than 90% donor-derived peripheral circulating immune cells 4 weeks after irradiation (see Figure E3A). Wild-type mice with Lrp5 null bone marrow–derived cells show no consistent differences in collagen abundance or tissue injury compared with Lrp5^+/+ bone marrow–reconstituted mice (see Figures E3B–E3G). After bleomycin administration, BAL fluid total protein was comparably elevated (3,312 ± 1,815 μg/ml, 4,319 ± 2,176 μg/ml), as well as soluble collagen content (1,563 ± 238 μg/ml, 1,644 ± 427 μg/ml) and total transforming growth factor (TGF)-β levels (407 ± 180 pg/ml, 482 ± 102 pg/ml). When we attempted to carry out the complementary chimeric approach using Lrp5^−/− mice as bone marrow recipients, we encountered increased lethality (∼50%), and those that survived were insufficiently chimeric (<88%). Given that Lrp5^−/− mice are known to manifest an osteoporotic phenotype (25), we reason that the Lrp5^−/− bone marrow niche is somehow compromised for receiving donor bone marrow–derived cells. Nevertheless, these data show that loss of Lrp5 from a bone marrow–derived cell type alone is insufficient to provide the protection observed in the Lrp5^−/− whole animal knock-out, indicating that Lrp5-mediated fibrosis depends, at least in part, on resident lung cell types. Evidence that the induction of Col1A1 mRNA in Lrp5^−/− bone marrow chimeras is significantly less than the Lrp5^+/+ chimeras (see Figure E3F) (2.1-fold ± 2.58 vs. 5.5-fold ± 2.63; P < 0.05) at Day 21 raises the possibility that Lrp5^−/− bone marrow cells may display an enhanced capacity to resolve fibrosis over time. Models to test this hypothesis are currently underway.

Lrp5 Loss Diminishes Abundance but Not Response to TGF-β

Wild-type Lrp5^+/+ mice manifest a clear elevation of both latent and active TGF-β levels (214.2 ± 107 pg/ml, 81.6 ± 42 pg/ml) in BAL fluid 21 days after bleomycin administration (Figure 5A). This increase in TGF-β is consistently observed in the bleomycin model, and the elevated TGF-β seems to be required and sufficient for the development of pulmonary fibrosis (26–28). Remarkably, bleomycin-injured Lrp5^−/− mice analyzed 21 days postinjury show substantially reduced levels of latent and active TGF-β (134.2 ± 46 pg/ml, 34.2 ± 14 pg/ml) compared with wild-type mice (Figure 5A).

Figure 5.

Lrp5 loss attenuates fibrosis through limiting transforming growth factor (TGF)-β abundance. (A) Scatter plot of latent and active TGF-β protein levels (ELISA method) in bronchoalveolar lavage from Lrp5^+/+ and Lrp5^−/− lungs harvested 21 days postinjury by saline (control, closed symbols) or bleomycin (open symbols). (B and C) TGF-β protein levels produced by alveolar type 2 (AT2) cells isolated from Lrp5^+/+ and Lrp5^−/− lungs harvested 6 (B) and 14 days (C) postinjury. (D and E) TGF-β protein levels produced by alveolar leukocytes isolated from Lrp5^+/+ and Lrp5^−/− lungs harvested 6 (D) and 14 days (E) postinjury. (F and G) Hematoxylin and eosin/trichrome staining of Lrp5^+/+ and Lrp5^−/− lungs harvested 21 days post adenovirus mediated–overexpression of TGF-β (F). Total soluble lung collagen harvested from mice 21 days postinfection was analyzed by Sircol method (G). Results are expressed as the mean TGF-β or collagen levels ± SEM from greater than or equal to four animals per condition. Each symbol represents data from a single mouse lung. *P < 0.05, **P < 0.01 by analysis of variance.

To determine whether this reduction in TGF-β protein levels is responsible for the protection, or simply a consequence of reduced fibrosis, we examined earlier time points after injury by bleomycin. Of interest, alveolar type 2 cells (AT2) isolated from Lrp5^−/− mice 6 days after bleomycin-injury produce less TGF-β compared with Lrp5^+/+ cells (1.44 ± 0.08 pg/μg total protein, 2.15 ± 0.15 pg/μg total protein) (Figure 5B). AT2 cells isolated from wild-type and null mice 14 days after bleomycin injury reveal no differences (0.98 ± 0.38 pg/μg total protein, 1.52 ± 0.41 pg/μg total protein) in TGF-β abundance (Figure 5C). Similarly, TGF-β production by alveolar leukocytes isolated from Lrp5^−/− mice is reduced compared with cells isolated from Lrp5^+/+ mice. Interestingly, this difference is observed 14 days (1.07 ± 0.52 pg/μg total protein vs. 2.53 ± 1.31 pg/μg total protein), but not 6 days (1.58 ± 1.31 pg/μg total protein vs. 1.44 ± 1.05 pg/μg total protein) after bleomycin injury (Figures 5D and 5E). These findings are consistent with the view that the bleomycin model of lung injury and fibrosis begins with alveolar epithelial injury within the first 6 days after bleomycin instillation and is followed by a fibrogenesis phase extending from Days 14 to 21 (29). These data also indicate that the loss of Lrp5 can blunt TGF-β production in both alveolar epithelial and immune cells with kinetics that follow the progression of this injury model.

The dependence of TGF-β protein abundance on Lrp5 is not apparently caused by an ability of LRP5 to intrinsically impact TGF-β signal transduction, because TGF-β promotes Smad2/3 activation by phosphorylation and Col1A1 induction similarly in Lrp5^+/+ and Lrp5^−/− fibroblasts (see Figure E4). The baseline difference in type 1 collagen protein levels is not simply caused by the loss of Lrp5, because a second independent isolation of Lrp5^+/+ and Lrp5^−/− fibroblasts revealed no differences in baseline collagen 1 expression. Even silencing of Lrp5 and its paralog, Lrp6, by siRNA in both an alveolar epithelial cell line and normal lung fibroblasts does not impact TGF-β activation of Smad (see Figure E5). These data indicate that Lrp5 loss impacts the abundance of, but not response to, TGF-β.

Mice Lacking Lrp5 Are Not Protected from TGF-β–induced Lung Fibrosis

If the reduction in TGF-β levels seen in Lrp5 null mice is indeed responsible for the attenuated fibrotic phenotype, then this protection should be overridden in a model of lung fibrosis driven by the direct delivery of active TGF-β to airways (26). Indeed, both wild-type and Lrp5 null mice demonstrate similar degrees of collagen induction (3,010 ± 864 μg/ml, 2,658 ± 486 μg/ml) when fibrosis is induced by adenovirus-mediated delivery of TGF-β (Figures 5F and 5G). These data show that restoration of full TGF-β production can reverse the protection seen in Lrp5 null mice, and suggest that TGF-β is a key, downstream mediator of Lrp5-driven fibrosis.

Discussion

In PBMC from a cohort of patients with IPF seen at the University of Pittsburgh, we find that Wnt signaling pathway is one of the top two overrepresented pathways associated with disease progression. Specifically, WNT3A, LRP6, and LRP5 are among the genes with the highest association with disease progression defined as a relative decline in 10% of FVC, relative decline in 15% of Dl_CO, or increase in 5% of the CPI. Importantly, LRP5 overexpression also correlates with disease severity at presentation in patients from the University of Chicago, indicating that LRP5 overexpression has prognostic value across distinct cohorts of patients with IPF. Unfortunately, it was not possible to determine whether LRP5 overexpression preceded disease progression in these patients, because serial pulmonary function test measurements were not available in the University of Chicago cohort. We further show that mice lacking the Wnt coreceptor, Lrp5, manifest reduced pulmonary fibrosis in the bleomycin-injury model. Using bone marrow–chimeric mice generated from mice lacking Lrp5, we show that Wnt/β-catenin signaling in the lung parenchyma is likely important for the development of fibrosis, because its role is not sufficient in inflammatory cells. These findings suggest that the changes in Wnt/β-catenin signaling observed in circulating mononuclear cells from patients may serve as a marker of poor prognosis, but alone are not causal in the development of fibrosis. Lastly, our findings suggest that Wnt/β-catenin signaling in the lung parenchyma enhances the abundance of TGF-β after bleomycin-induced lung injury. In the presence of high levels of active TGF-β whose expression is driven by a viral promoter, the attenuation of Wnt/β-catenin signaling by Lrp5 loss has no effect on the development of lung fibrosis. Together, these results suggest that inhibitors of the Wnt pathway might be more effective in patients with early fibrosis, before the autonomous activation of TGF-β in the lung is established (30–33).

Although elevated expression of WNT3A, LRP5, and LRP6 seems to predict IPF progression, it is worth noting, however, that overexpression of the Wnt pathway inhibitor, DKK1, also correlates with disease progression (Figure 1A). Evidence that overexpression of “activators” and “inhibitors” of Wnt signaling can be associated with disease progression underscores the difficulty in inferring whether “too much” or “too little” signaling is the problem in fibrosis from expression data alone (particularly when expression analysis is based on a mixed cell population, such as PBMCs). Nonetheless, a strength of the present study is our combined use of an animal model where Wnt/β-catenin signaling is known to be diminished (e.g., Lrp5 null model [20]) along with gene expression and outcomes data collected from human patients with IPF.

Our finding that mice globally deficient in Lrp5 are protected against bleomycin-induced fibrosis is consistent with previous studies showing that enhanced Wnt/β-catenin signaling can drive pulmonary fibrosis (reviewed in [34]). However, ours is the first study that uses a genetic loss-of-function approach targeting a Wnt pathway component that is mutated in other human diseases (35–39). The Lrp5 whole animal knock-out (40), which has no observable phenotype in the lung in the absence of stress, may be better suited to modeling human lung disease than mice deficient in β-catenin, which may impact both cell-cell adhesive and Wnt signaling functions (41).

The manner in which Lrp5 loss attenuates bleomycin-induced pulmonary fibrosis is likely through a reduction in β-catenin signaling activity. Fibroblasts isolated from Lrp5-deficient mice have a blunted response to Wnt3a, but respond normally to a degradation-resistant form of β-catenin, which activates signaling downstream of LRP5/6 receptors. In addition, a small molecule inhibitor of β-catenin/T-cell factor transcriptional activity, iCRT3 (23), mimics the protective phenotype seen by loss of Lrp5, and adds to a growing list of Wnt pathway inhibitors that attenuate fibrosis in the bleomycin model (8, 9, 11). Whether iCRT3 attenuates β-catenin binding to other transcription factors (e.g., HIF1α, Foxm1) is not yet known.

Although Lrp5 is best known for its role as a Wnt coreceptor, it is formally possible that its loss could attenuate fibrosis through a β-catenin signaling–independent mechanism. In this regard, Lrp5 knock-out mice show elevated expression of tryptophan hydroxylase 1 in a manner that seems independent of β-catenin (42). Tryptophan hydroxylase 1 is a rate-limiting enzyme in the synthesis of serotonin, and high tryptophan hydroxylase 1 levels lead to systemic elevation of serotonin and bone loss in Lrp5^−/− mice (42). However, because serotonin and its receptors are found to be increased in lung fibrosis, where their inhibition limits fibrosis in the bleomycin model (43, 44), the reduced fibrosis we observe in Lrp5 null mice is likely independent of an LRP5/serotonin signaling axis. Lastly, recent data show that the Lrp5/6 inhibitor, Dkk-1, can blunt expression of fibrogenic markers in pericytes through broadly inhibiting platelet-derived growth factor–, TGF-β–, and connective tissue growth factor–activated MAPK and JNK signaling cascades independently of β-catenin (45). However, our evidence that primary lung fibroblasts lacking Lrp5 (see Figure E4) or knock-down of Lrp5 and 6 in both a lung epithelia cell line and normal lung fibroblasts show no difference in the ability of TGF-β to promote Smad phosphorylation (see Figure E5) raises the possibility that the inhibitory effects of DKK-1 may target cell surface receptors in addition to Lrp5 and 6.

Cross-talk between the Wnt/β-catenin and TGF-β pathways has been the subject of numerous studies, where evidence places Wnt/β-catenin signaling both downstream and upstream of TGF signals. For example, β-catenin signaling can up-regulate TGF-β message in some but not all contexts (46, 47), and TGF-β signaling can lead to β-catenin signaling through Wnt-dependent (10) and-independent mechanisms (48). In our study, alveolar epithelial type 2 cells and leukocytes isolated from Lrp5^−/− mice manifest reduced capacity to produce TGF-β protein compared with cells isolated from Lrp5^+/+ mice. Because Lrp5^−/− lung fibroblasts and alveolar epithelial cells silenced for both Lrp5 and 6 are similarly responsive to exogenous TGF-β (see Figures E4 and E5), it may be that injury-induced expression of other cytokines may be enhancing Lrp5/β-catenin–mediated expression of TGF-β (Figure 6). Curiously, the requirement for Lrp5 in TGF-β production in AT2 cells is observed 6 but not 14 days post bleomycin injury. Conversely, the requirement for Lrp5 in TGF-β production in leukocytes is observed 14 but not 6 days postinjury (Figure 5). In the bleomycin model, alveolar macrophages are thought to produce most TGF-β (49), and data suggest that production of TGF-β in macrophages in an IL-13 overexpression transgenic model can drive fibrosis (50). However, evidence that Lrp5^−/− bone marrow–derived cells are not sufficient to provide the protection observed in a total Lrp5^−/− mouse indicates that the lung parenchyma also contributes profibrotic LRP5 activity (see Figure E3). Because secretion of TGF-β by alveolar epithelial cells is seen in both IPF and early time points of the bleomycin-injury model, where it is considered to be an important contributor to fibrosis (51–53), we reason that the protection from bleomycin-mediated pulmonary fibrosis in Lrp5 null mice may be mediated by reductions in TGF-β production by both Lrp5 null AT2 cells and macrophages. Of interest, the contribution of Lrp5 to TGF-β production follows the temporal progression of this injury model, from AT2 cell injury (Days_0–6) to immune cell involvement (Day₁₄), consistent with the view that epithelial-mesenchymal cross-talk involving both Wnt and TGF-β signaling relationships are drivers of fibrosis. Lastly, evidence that Lrp5 null mice are not protected in the TGF-β–induced model of fibrosis (Figures 5F and 5G) strongly suggests that the reduction in TGF-β levels is a means through which Lrp5 null mice are protected in the bleomycin model.

Figure 6.

Model for Lrp5 contribution to transforming growth factor (TGF)-β production. Lrp5 loss in mouse alveolar epithelial cells and leukocytes (presumably macrophages) leads to a reduced capacity to produce TGF-β protein (Figure 5). Presence of Lrp5 is shown in A; its absence is shown in B. These data raise the possibility that Lrp5 contributes to TGF-β gene expression, possibly through T-cell factor (TCF)–binding elements (TBE) in the TGF-β-promoter/enhancers that contribute to TGF-β expression. In alveolar epithelial cells and fibroblasts, neither Lrp5 nor Lrp6 are required for the actual transduction of TGF-β signals, as quantified by abundance of phosphorylated Smad2 (see Figures E4 and E5). Lrp5 loss also does not impact TGF-β–dependent up-regulation of TGF-β (see Figures E4B and E4C), raising the possibility that the Smad-binding element (SBE) can contribute more strongly to TGF-β expression than the Lrp5/β-catenin signaling mechanism.

Our evidence that TGF-β–driven lung fibrosis is not impacted by the loss of Lrp5 has implications for the success of Wnt pathway inhibitors in the treatment of fibrosis. However, we should point out that work by the Distler group finds that the Wnt pathway inhibitor, Dkk-1, seems to be a robust inhibitor of fibrosis driven by TGF-β and other means (10). There are two ways to reconcile our findings with those of the Distler group. First, our study addresses the consequence of attenuating β-catenin signaling to TGF-β–driven fibrosis through genetic deletion of Lrp5, but not its paralog, Lrp6, whereas Akhmetshima and coworkers (54) used a Dkk-1 transgenic mouse, which should inhibit Wnt signaling via both LRP5 and 6 receptors. Thus, it stands to reason that the Dkk-1 transgenic experiment may inhibit β-catenin signaling more effectively, because it can inhibit signaling via both LRP5 and 6. The Dkk-1 transgenic mouse model may also be more potent because of the level of Dkk-1 overexpression in this mouse. A second way to reconcile our data is to consider evidence from the Duffield group showing that recombinant Dkk-1 protein can broadly antagonize profibrotic signaling via platelet-derived growth factor and TGF-β, apparently independently of β-catenin (45). Thus altogether, these data suggest that only robust β-catenin signaling inhibition can reduce TGF-β–driven fibrosis, or that Dkk-1 is such an effective inhibitor of TGF-β–driven fibrosis because of its ability to inhibit fibrogenic activities that are both β-catenin dependent and independent.

Nonetheless, several studies provide evidence that small molecule inhibitors of β-catenin signaling (8–10) or antibodies directed to an extracellular target of β-catenin signaling (11) can attenuate fibrosis in the bleomycin model, suggesting not only that elevated β-catenin signaling drives fibrosis, but that inhibiting this pathway is ameliorative in this model. However, it is important to recognize that targeted loss of β-catenin in AT2 cells using a surfactant protein C-Cre recombinase can worsen bleomycin-mediated fibrosis (55), possibly because forced inhibition of β-catenin signaling can reduce AT2 cell survival and migration in culture (55–57), which may exacerbate epithelial repair processes in vivo. These latter studies indicate that the contribution of Wnt/β-catenin signaling to tissue repair is complex, with cell-type and signaling intensity-dependent roles that need to be considered when targeting this pathway therapeutically (41). Gene expression profiling in PBMC or markers of TGF-β activation in the lung might identify subgroups of patients with IPF most likely to benefit from Wnt pathway inhibitors (58).