Article | Published:

Thyroid hormone receptor repression is linked to type I pneumocyte–associated respiratory distress syndrome

Nature Medicine volume 17, pages 14661472 (2011) | Download Citation

Abstract

Although the lung is a defining feature of air-breathing animals, the pathway controlling the formation of type I pneumocytes, the cells that mediate gas exchange, is poorly understood. In contrast, the glucocorticoid receptor and its cognate ligand have long been known to promote type II pneumocyte maturation; prenatal administration of glucocorticoids is commonly used to attenuate the severity of infant respiratory distress syndrome (RDS). Here we show that knock-in mutations of the nuclear co-repressor SMRT (silencing mediator of retinoid and thyroid hormone receptors) in C57BL/6 mice (SMRTmRID) produces a previously unidentified respiratory distress syndrome caused by prematurity of the type I pneumocyte. Though unresponsive to glucocorticoids, treatment with anti–thyroid hormone drugs (propylthiouracil or methimazole) completely rescues SMRT-induced RDS, suggesting an unrecognized and essential role for the thyroid hormone receptor (TR) in lung development. We show that TR and SMRT control type I pneumocyte differentiation through Klf2, which, in turn, seems to directly activate the type I pneumocyte gene program. Conversely, mice without lung Klf2 lack mature type I pneumocytes and die shortly after birth, closely recapitulating the SMRTmRID phenotype. These results identify TR as a second nuclear receptor involved in lung development, specifically type I pneumocyte differentiation, and suggest a possible new type of therapeutic option in the treatment of RDS that is unresponsive to glucocorticoids.

Main

At birth, the neonate undergoes a profound metabolic transition and has a dramatically enhanced dependence on postnatal oxidative metabolism. Although this metabolic switch is a hallmark of the transition from fetus to newborn, little is known about the molecular genetics that direct this process. Physiologically, it is the first breath that expands the lung, enabling enhanced oxygenation of the blood. Failure of the lung to expand, such as in infant RDS, is one of the most common causes of neonatal mortality. Lung morphogenesis begins with the embryonic lung buds originating from the foregut endoderm to form airways that branch into the millions of alveoli required for air exchange immediately after birth1,2. These alveoli are lined by two types of epithelial cells: gas-permeable type I pneumocytes, which are responsible for air exchange, and type II pneumocytes, which produce surfactant to reduce surface tension3,4,5.

Nuclear receptors are ligand-activated transcription factors that have key roles in both normal physiology and pathological conditions6. Their activities are not only determined by the availability of their respective ligands but also by their association with certain coactivators and co-repressors. SMRT was initially identified as a transcriptional co-repressor that maintains the 'off state' of non-liganded nuclear receptors7,8. Its interaction with nuclear receptors is mediated by two C-terminal Leu-X-X-X-Ile-X-X-X-Ile/Leu motifs termed receptor interacting domains (RIDs)9,10. SMRT facilitates transcriptional repression by serving as a scaffold protein to recruit histone deacetylase complexes and chromatin remodeling factors11,12,13.

We generated SMRT knock-in mice (SMRTmRID) with point mutations in both the RID1 and RID2 domains that specifically disrupt the interaction between SMRT and nuclear receptors14. Here we show that these mice on a pure C57BL/6J background die shortly after birth from acute RDS resulting from an abnormal terminal differentiation of type I pneumocytes. The lungs of these mice had normal type II pneumocyte morphologies, including surfactant production, and showed an unaltered gene expression program. Thus, these SMRTmRID mice provide evidence for a unique nuclear receptor–dependent pathway in type I pneumocyte differentiation. Notably, propylthiouracil and methimazole, two antithyroid drugs, but no other nuclear receptor antagonists, restored functional type I pneumocytes, fully rescuing lung development and viability of the SMRTmRID mice. Furthermore, we provide evidence that Klf2 is downstream of TR-SMRT signaling and is both necessary and sufficient for type I pneumocyte differentiation.

Prenatal administration of glucocorticoids has long been known to promote type II pneumocyte maturation and is clinically used to mitigate the severity of infant RDS. This study identifies a second crucial nuclear receptor pathway in lung development that is controlled by TR, SMRT and Klf2, and when disabled, blocks type I pneumocyte differentiation, resulting in catastrophic lung collapse even in the presence of surfactants. The ability to fully rescue the syndrome with a commonly used drug suggests a potential new avenue in the management of lung prematurity and infant respiratory distress.

Results

SMRTmRID mice die from acute respiratory failure

We previously described SMRTmRID mice with point mutations in the Ncor2 (also known as SMRT) gene that specifically disrupt SMRT interactions with nuclear receptors14. Although these mice are viable in a C57BL/6-Sv129 mixed background, only 14% (49 from a total of 361 pups) survived weaning (Supplementary Fig. 1). Viability dropped to less than 1% when we backcrossed these mice to a pure C57BL6/J background (>99.5% BL6/J) (4 out of 494 pups) (Supplementary Fig. 1). This drop in viability was not because of early embryonic lethality, as all genotypes were represented at close to a Mendelian ratio at mid-gestation (embryonic day 13.5 (E13.5); Supplementary Fig. 1).

The cause of this postnatal lethality seems to be a form of acute respiratory distress marked by severe dyspnea, cyanosis (Fig. 1a), deoxygenated blood (Fig. 1b) and diffuse lung atelectasis, as shown by lungs that sink readily in saline (Fig. 1c). Gross evaluations showed that these lungs had normal orientation and lobation; there were no obvious upper respiratory airway obstructions or diaphragmatic defects. Analyses of whole-body sections at different planes also revealed no histological abnormalities except in the lung (data not shown).

Figure 1: SMRTmRID mice die postnatally from acute respiratory failure.
Figure 1

(a) A representative newborn SMRT litter with two SMRTmRID (mRID) pups shown in the middle. Arrows indicate the visible lungs. Het, heterozygous of SMRTmRID allele. (b) A representative picture of blood from newborn WT and SMRTmRID littermates. (c) Comparison of the lungs from newborn WT and SMRTmRID littermates in PBS. (d) Microscopic images of H&E-stained lung tissue from newborn SMRT littermates and the quantification of alveoli sizes. Scale bars in the top images, 200 μm. Scale bars in the bottom images, 50 μm. Error bars, s.e.m.

Microscopically, lungs from both wild-type (WT) and SMRTmRID mice from E18.5 to postnatal day 0 (P0) were in the saccular stage (Fig. 1d and Supplementary Fig. 2). Lungs from SMRTmRID mice appeared partially collapsed and had irregular and narrower airspaces. Alveoli in these lungs were about eight times smaller and the alveolar walls were thickened, hypercellular and lined by cuboidal cells compared to lungs from WT mice. Other pulmonary structures appeared histologically normal. We observed no prenatal gross or histological differences between SMRTmRID and WT mice at the early (E13.5) and late (E16.5) pseudoglandular stages, pointing to a late-stage maturation problem (Supplementary Fig. 2 and data not shown).

SMRTmRID lungs retain premature type I pneumocytes

To determine the cause of postnatal death in SMRTmRID mice, we next examined their lungs under higher magnification using both light (Fig. 2a) and electron (Fig. 2b) microscopy. Although we readily observed type II pneumocytes showing normal cuboidal morphology, we found few cells with the distinctive flattened morphology typical of type I pneumocytes in SMRTmRID alveoli (Fig. 2a). We made similar observations at E18.5 (not shown). Under electron microsopy analysis, the alveoli in SMRTmRID mice contained normal amounts of surfactant, lamellar bodies and tubular myelin. However, while the alveolar walls in the mutant lungs were lined predominantly by type II pneumocytes of normal morphology, there were very few type I pneumocytes with the characteristic thin, gas-permeable cytoplasmic extensions that cover the capillaries (Fig. 2b).

Figure 2: Lungs of SMRTmRID mice lack mature type I pneumocytes.
Figure 2

(a) High-magnification light microscopic images of H&E-stained lung tissue from newborn WT and SMRTmRID littermates. Arrowheads indicate some of the type II pneumocyte–like cells. Scale bars, 20 μm. (b) Electron microscopic images of lung tissue from newborn WT and SMRTmRID littermates. Scale bars, 5 μm. (c) Immunostaining of type I pneumocyte markers (T1α and Cav1) in E18.5 WT and SMRTmRID littermate lungs. Scale bars, 50 μm. (d) Expression of type I and II pneumocyte markers in E17.5 WT and SMRTmRID littermate lungs. We analyzed gene expression in each individual embryo. *P < 0.05, ***P < 0.001. NS, not significant. Error bars, s.e.m.

Although immunohistochemical examination of markers specific for both type I and type II cells revealed no changes in surfactant protein levels (at E18.5; Supplementary Fig. 3), we observed a dramatic reduction of typical type I pneumocyte markers (T1α and Cav1) in lungs from SMRTmRID (Fig. 2c). As determined by the marker staining, type I pneumocyte cytoplasmic extensions cover only 10–20% of the alveolar surface in SMRTmRID mice compared to greater than 90% coverage in WT littermates. Though less marked, the mRNA levels of type I cell markers (T1α, Aqp5 and Cav1) were also significantly lower at both E17.5 and E18.5 (Fig. 2d and Supplementary Fig. 4), suggesting that transcription of the type I pneumocyte differentiation program is compromised at these stages. Consistent with the immunohistochemistry results, mRNA levels of type II cell marker surfactant proteins were unaffected. The mRNA levels of the transcription factors Nkx2-1 (also known as Ttf1 or T/EBP), Foxa2 (also known as HNF3β) and Gata6, which were previously implicated in lung morphogenesis15,16,17, were similar between WT and SMRTmRID mice (Fig. 2d and Supplementary Fig. 4). Because glycogen is used in type II cells to generate surfactants, we performed periodic acid–Schiff staining, which revealed similar glycogen levels in WT and SMRTmRID lungs, further confirming that type II cells were functionally intact (Supplementary Fig. 5). Finally, as pneumocyte and alveolar development depend on normal blood flow and vascular architecture, we analyzed the terminal vascular differentiation marker Pecam1. This analysis revealed normal alveolar vasculogenesis within the developing alveolar structure (data not shown), which, in combination with the above data, suggests that type II pneumocytes, alveolar structure and vasculogenesis are normal in SMRTmRID mice.

Antithyroid drugs rescue type I pneumocytes and SMRTmRID mice

The above observations led to the hypothesis that SMRT-dependent repression of one or more nuclear receptors is required to establish the type I pneumocyte lineage and prevent acute respiratory failure. As SMRTmRID cells show higher basal receptor activity as well as an enhanced response to hormones such as perixosome proliferator-activated receptor γ (PPAR-γ) agonists14 and triiodothyronine (T3) (Supplementary Fig. 6), we conjectured that inhibiting nuclear receptor activity might rescue the SMRTmRID phenotype.

Accordingly, we tested several nuclear receptor antagonists by delivery at different doses to pregnant females at E16.5, which is just before the gestational age when we observed the differences in gene expression and morphology. Among these antagonists, those of retinoic acid receptor α (RARα) (RO 41-5253), estrogen receptors (tamoxifen), and glucocorticoid receptor and progesterone receptor (RU486) did not affect the survival rate of SMRTmRID pups (Supplementary Table 1). Although glucocorticoids are used clinically in infant RDS to promote surfactant production from type II pneumocytes, dexamethasone was not able to rescue SMRTmRID pups, consistent with a type I pneumocyte defect. In contrast, we observed that two different clinically approved anti-thyroid drugs that block iodine incorporation into thyroid hormone rescued the SMRTmRID pups (Supplementary Table 1). Both oral propylthiouracil (PTU, which also inhibits conversion of thyroxine (T4) to T3) and methimazole (MMI) administered through intraperitoneal injection were able to achieve efficient rescue.

Not only were PTU- and MMI-treated SMRTmRID mice born alive, but, notably, they showed no evidence of respiratory distress and survived to adulthood. Upon gross examination, lungs from these mice were pink, inflated and floated in saline (Fig. 3 compared to Fig. 1a,c). We found blood from these mice to be properly oxygenated (Fig. 3b compared to Fig. 1b). Microscopically, both WT and SMRTmRID lungs from PTU-treated mice were normal and in the saccular stage at E18.5 and P0 (Fig. 3d). Electron microscopy revealed terminal sacculi that were composed of normal septae with continuous cytoplasmic extension coverage from flattened type I pneumocytes as well as cuboidal type II cells containing lamellar bodies (Fig. 3e). Consistent with morphological rescue, the transcriptional activation of type I pneumocyte markers was restored, and there were no significant differences in type I or II cell markers or in lung transcription factor expression between PTU-treated WT and SMRTmRID mice (Supplementary Fig. 4).

Figure 3: Antithyroid drugs rescue SMRTmRID mice and restore type I pneumocyte development.
Figure 3

(a) A representative PTU-treated newborn SMRT litter. Arrows indicate the visible lungs. (b) A representative picture of the blood from PTU-treated newborn WT and SMRTmRID littermates. (c) Comparison of the lungs from PTU-treated newborn WT and SMRTmRID littermates in PBS. (d) Microscopic images of H&E-stained lung tissue from WT, SMRTmRID and PTU-treated SMRTmRID littermates (E18.5 and P0). Scale bars, 50 μm. (e) Electron microscopic images of lung tissue from PTU-treated newborn WT and SMRTmRID littermates. Scale bars, 5 μm.

To confirm that PTU decreases serum thyroid hormone levels in treated litters, we developed a very sensitive mass spectrometry assay to measure the concentrations of the thyroid hormones T4 and T3 and reverse T3 (rT3) from small blood volumes (10–15 μl). The results showed that the serum T4 levels in both genotypes were reduced after a few days of PTU treatment; we did not detect serum T3 (Supplementary Fig. 7 and Supplementary Table 2). It is notable that serum T4 concentrations in SMRTmRID mice were lower than in their WT littermates. This is probably caused by higher thyroid hormone receptor activity (as observed in SMRTmRID mouse embryonic fibroblasts; Supplementary Fig. 6) in their hypothalamic-pituitary-thyroid axes, resulting in suppressed T4 production. We also note that the major embryonic form of thyroid hormone is rT3, a T4 metabolite that does not activate the thyroid hormone receptor. In contrast, and although thyroid hormone crosses the placenta, T4 is the dominant form of thyroid hormone in pregnant females, suggesting that at this stage of embryonic development there is an active mechanism to degrade T4 to rT3 to lower thyroid hormone receptor signaling in pups18 (data not shown). TR is encoded by two genes (Thra (also known as Trα) and Thrb (also known as Trβ)) with distinct expression patterns; knockout animal studies have suggested important yet different physiological functions of these two genes19,20,21,22,23. Quantitative RT-PCR (qRT-PCR) analysis of E17.5 to P0 lungs revealed that both TRα and TRβ are expressed at these stages (Supplementary Fig. 4 and data not shown), with TRβ likely to be the more abundant form (qPCR Ct values of 25 for Thrb and 28 for Thra).

We also tested the effect of the timing of PTU treatment on survival rate. Notably, PTU rescued the SMRTmRID mice only when started at E16.5 and withdrawn right after birth, coinciding with dramatic lung differentiation and branching from the pseudoglandular stage to the saccular-alveolar stage (data not shown). Shorter periods of treatment failed to effect rescue, and longer periods of treatment (starting PTU before E16.5 or withdrawing after P0) resulted in the death of all pups regardless of their genotype, presumably caused by cretinism24.

SMRT regulates Klf2 and type I pneumocyte gene program

The above results strongly indicated that TR and SMRT signaling regulate type I pneumocyte formation without affecting other cell types in developing lungs. As little is known about this cell lineage5,15, the SMRTmRID mouse provides a unique model to probe the molecular mechanisms of type I pneumocyte differentiation. As TR and SMRT do not directly bind the promoter regions of Pdpn (also known as T1α), Aqp5 and Cav1 (data not shown), we hypothesized that one or more 'mediator' factors downstream of the thyroid hormone receptor and SMRT must determine type I pneumocyte differentiation.

To identify such mediator proteins, we next performed genome-wide gene expression analysis by microarray of pooled WT and SMRTmRID lungs at E18.5 with or without PTU treatment. As type I pneumocytes comprise only a small percentage of total lung cells, we observed only 21 robustly expressed genes to be markedly changed between WT and SMRTmRID mice and remained statistically unaffected by PTU rescue at E18.5 (Supplementary Table 3). We compared expression levels by real-time PCR of these 21 candidate genes in individual littermate-controlled WT and SMRTmRID embryos at both E17.5 and E18.5 when type I markers were reduced and found that only one gene, Klf2, also known as Lklf (encoding the lung Kruppel-like factor), was significantly different at both time points and remained statistically unchanged with PTU treatment (Fig. 4a and data not shown). More importantly, Klf2 protein levels were almost undetectable by immunohistochemistry in lungs from SMRTmRID mice, whereas strong nuclear staining of Klf2 could be clearly found in WT branching saccular epithelial cells (Fig. 4b).

Figure 4: Klf2 is reduced in SMRTmRID lungs and activates the type I pneumocyte gene program.
Figure 4

(a) Expression of Klf2 in E17.5 and E18.5 WT and SMRTmRID littermate lungs. We analyzed gene expression in each individual embryo. *P < 0.05, **P < 0.01. (b) Immunostaining of Klf2 in E18.5 and P0 WT and SMRTmRID littermate lungs. Scale bars, 50 μm. (c) Adenoviral-mediated Klf2 expression induces type I pneumocyte marker expression in MLE-12 cells. The multiplicity of infection (MOI) of the adenovirus used is indicated. *P < 0.05, ***P < 0.001. AdGFP, adenovirus containing GFP; AdKlf2, adenovirus containing Klf2. (d) Klf2 binds to the promoter region of type I pneumocyte marker genes in a ChIP assay. *P < 0.05, **P < 0.01. (e) Klf2, but not Nkx2-1, directly activates Pdpn (T1α), Aqp5 and Cav1 promoters (mT1α, mAqp5 and mCav1). We performed transient transfections in triplicate wells in CV-1 cells and repeated these transfections at least three times. Luciferase values normalized to β-galactosidase are shown. **P < 0.01. Error bars, s.e.m. pAdempty, pAdklf2 and pAdNkx2-1 are adenovirus vectors. NS, not significant.

We next investigated whether TR and SMRT could repress Klf2 transcription through direct binding to its promoter or enhancers. Use of chromatin immunoprecipitation (ChIP) followed by deep sequencing to map genome-wide SMRT binding sites in macrophages (ref. 25 and G.B. and R.M.E., unpublished data) revealed that SMRT binds strongly to a conserved enhancer region downstream of the mouse Klf2 locus (Supplementary Fig. 8). Conventional ChIP confirmed that both TR and SMRT also bind to this conserved Klf2 enhancer region (Supplementary Fig. 9) in MLE-12 cells.

To determine if this enhancer contains a positive or negative thyoid hormone response element (TRE), we cloned it into the 3′ region of a pGL4 luciferase reporter vector. Transient transfection assays with a classical DR4 (TRE) showed that T3 increases the DR4-luciferase activity in the presence of TRβ and RXRα (Supplementary Fig. 10). However, the Klf2 enhancer is repressed by T3, identifying it as a negative TRE. Negative TREs, such as in the Tsh promoter, are repressed by T3 and activated in its absence, similar to what is seen in Klf2.

Klf2 is expressed predominantly in the lung throughout embryonic and adult stages26. We next asked whether Klf2 is sufficient to induce a type I pneumocyte gene program. To this end, we overexpressed Klf2 in a mouse lung carcinoma cell line, MLE-12, using adenoviral vectors. The dose of adenovirus we used resulted in approximately 30–50% cell infection but did not affect cell survival or growth. Expression of type I markers, including Pdpn (T1α) and Aqp5, was induced by ectopic Klf2 in a dose-dependent manner (Fig. 4c). Typical type II pneumocyte markers, specifically surfactant mRNAs, remained either unchanged (Sftpa1 (also known as Sp-a) and Sftpb (also known as Sp-b)) or slightly increased (Sftpc (also known as Sp-c)) by Klf2 overexpression (data not shown), suggesting that Klf2 specifically drove the type I cell differentiation program.

As Klf2 is a transcription factor, we next tested whether Klf2 could directly bind and activate type I cell marker gene promoters. We noticed that proximal promoters of multiple type I cell markers contain conserved Klf2 binding sites. ChIP assays showed that Klf2 directly binds to the promoter regions containing these sites (Fig. 4d). Transient transfection assays showed that Klf2 directly activated the promoters of these type I markers, including Pdpn (T1α), Aqp5 and Cav1 (Fig. 4e), whereas the well established type II pneumocyte transcription factor Nkx2-1 had little effect. These results suggest that Klf2 is sufficient to directly activate a type I pneumocyte gene program in vitro.

Klf2 is essential for type I pneumocyte differentiation

We next asked whether Klf2 is required for type I pneumocyte differentiation and normal lung development in vivo. However, Klf2 knockout mice die between E11.5 and E14.5 (ref. 27,28), precluding the analysis of their lungs at later stages of development (from E16.5 to P0). To circumvent this early embryonic lethality, we generated chimeric mice using Klf2−/− embryonic stem (ES) cells to test whether Klf2 is required for normal lung development. Among 38 chimeras that we generated from eight litters, we found two pups that showed signs of respiratory distress and died soon after delivery by Caesarean section at E18.5. These two pups (numbers 19 and 33) had cyanosis with uninflated lungs (Fig. 5a), closely phenocopying the RDS of the SMRTmRID pups. Viable pups had no or low chimerism, whereas the two distressed pups had highly chimeric Klf2−/− lungs (data not shown), resulting in low Klf2 mRNA and protein levels compared to their littermates (Fig. 5b,c). We also found these two mice to have normal expression levels of Klf2 in the other tissues we examined (tail, liver and muscle; data not shown), suggesting that the phenotype is caused by the high chimerism in the lungs that is not present in other tissues. Expression of other lung transcription factors including Nkx2-1, Gata6 and Foxa2 (HNF3β) was essentially unchanged (Supplementary Fig. 11).

Figure 5: Klf2 is essential for type I pneumocyte and normal lung development in vivo.
Figure 5

(a) A representative picture showing a highly Klf2-chimeric mouse (mouse number 33, shown in the middle) and its sinking lung in PBS compared to its littermates. The lungs of the littermates are indicated by arrows. (b) Expression of Klf2 and type I pneumocyte markers in highly Klf2-chimeric mice and their respective littermates. Error bars, s.e.m. (c) Immunostaining of Klf2 and type I pneumocyte markers in highly Klf2-chimeric mice and their respective littermates. Scale bars, 50 μm. (d) H&E-staining of lungs from highly Klf2-chimeric mice and their respective littermates. Scale bars, 100 μm. (e) A working model proposing that the glucocorticoid receptor (GR) and thyroid hormone receptor (TR) regulate distinct pathways of pneumocyte development.

Histological examination revealed that the two highly chimeric pups had lung atelectasis (Fig. 5d), with few flattened cells typical of type I pneumocytes, whereas they had abundant numbers of cuboidal type II–like cells. All type I markers were expressed at much lower levels in these two pups (Fig. 5b). Immunohistochemistry revealed a marked reduction of type I marker proteins in their lungs (Fig. 5c). The type II pneumocytes and endothelial cells were not affected, as determined by Sp-b, Sp-c and Pecam1 staining (Supplementary Fig. 12 and data not shown). These results show that Klf2 is essential for type I pneumocyte differentiation and normal lung development in vivo.

Discussion

Embryonic lung development is controlled by a precisely regulated cascade of morphogens, cellular signaling molecules, hormones and transcription factors15,16,29,30,31,32. Specific cell lineage determination factors have been characterized for many specialized cells of the lung17,33,34,35,36,37. However, the factors that direct the terminal differentiation of type I pneumocytes, which are the cells responsible for gas exchange in the lung, have not been identified5,15. In this study, we delineate a previously unidentified signaling pathway in which a TR–SMRT complex is required for type I pneumocyte terminal differentiation and provide evidence that the transcription factor Klf2 is downstream of TR signaling in normal lung morphogenesis. We show that Klf2 activates a genetic program consistent with type I—but not type II—pneumocyte development through direct interactions with type I marker promoters. Notably, mice without lung Klf2 lack mature type I pneumocytes and die postnatally, closely recapitulating the phenotype of SMRTmRID mice. These findings identify a crucial developmental role for thyroid hormone in the maturation of type I pneumocytes and suggest that Klf2 is an essential mediator of thyroid hormone signaling in the lung.

Nuclear receptors and their co-regulators have important roles in many physiological and pathological conditions. In the developing lung, the glucocorticoid receptor is well recognized for its role in promoting surfactant production by type II pnumocytes. The retinoid acid receptors are crucial for early lung morphogenesis and tracheal-esophageal separation38. However, the role of the thyroid hormone receptor is less clear. Although some earlier reports have shown that thyroid hormone enhanced the synthesis of surfactant phospholipid from type II pneumocytes, most studies have shown that thyroid hormone does not enhance surfactant protein transcription and synthesis (summarized in ref. 39). In support of this, large-scale human clinical trials have shown that antenatal administration of thyrotropin-releasing hormone in addition to surfactant supplement treatment does not provide further benefit to the neonatal outcome of preterm infants40. We show here that antithyroid hormone drugs could rescue the isolated type I pneumocyte defects in SMRTmRID mice. We also show that TR achieves this rescue at least partially through direct regulation of Klf2 transcription. These results point to a crucial and previously unappreciated role of TR in type I pneumocyte development. However, it remains possible that mutations affecting SMRT may alter the type I pneumocyte differentiation program partially through another orphan nuclear receptor or an unknown transcriptional partner, or the effects of antithyroid hormone drugs could be mediated though a TR–independent mechanism. It is also likely that in addition to directly regulating Klf2, the TR–SMRT complex can determine type I cell fate by affecting the proliferation and/or apoptosis of its precursors. Future work using animals that have altered TR signaling in the developing lung will be instrumental in proving the essential role of TR in type I pneumocyte development.

Given the important role of glucocorticoid receptor signaling in lung development and surfactant expression, antenatal glucocorticoid therapy has been successfully used to accelerate lung development in premature infants and to ameliorate infant RDS41,42,43. It is notable that approximately 20–30% of infants with RDS do not respond to surfactant replacement therapy, and babies with lung hypoplasia who are extremely premature (less than 24 weeks of gestation in humans) do not respond well to exogenous surfactant replacement because of structural immaturity44, suggesting that overcoming lung immaturities in premature infants requires the full function of both types of pneumocytes. The rescue of type I pneumocyte immaturity in SMRTmRID mice by maternal antithyroid hormone drug treatment (PTU or MMI) identifies a key and previously unknown role of the thyroid hormone receptor in controlling this crucial transition of the fetus associated with the first breath.

In summary, we show that SMRT has a nonredundant role to the nuclear receptor co-repressor (NCoR) and that its association with the thyroid hormone receptor is crucial for the terminal differentiation of the type I pneumocyte. Although the glucocorticoid receptor has long been known to promote type II pneumocyte development, our findings reveal a previously unidentified thyroid hormone receptor–dependent pathway for type I cell maturation and lung development (Fig. 5e).

Methods

Mouse experiments.

All mouse procedures were approved by and carried out under the guidelines of the Institutional Animal Care and Use Committee of the Salk Institute. All mice were maintained in a temperature- and light-controlled environment and received a standard diet (PMI laboratory rodent diet 5001, Harlan Teklad) unless otherwise noted. SMRTmRID mice were generated as described14 and were backcrossed to a C57BL6/J background for at least seven generations. SMRT heterozygous × heterozygous breeding pairs were checked daily at 7:00 a.m. for vaginal plugs. We considered 12:00 p.m. of the day that the plug was observed to be embryonic day 0.5 (E0.5). In our facility, it took mice in a pure C57BL6 background 19 d to develop from conception to natural birth (P0 to E19). The doses and timing of the drugs used in the timed pregnant female mice are listed in Supplementary Table 1. The diet containing 0.15% PTU and no iodine was from Harlan Teklad (TD 97061), RO 41-5253 was from Biomol and all other chemicals were from Sigma.

Gene expression analysis.

We isolated total RNA from mouse tissues or cells using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. We synthesized cDNA from 1 μg of DNase-treated total RNA using Superscript II Reverse Transcriptase (Invitrogen). We quantified mRNA levels by real-time qRT-PCR using SYBR Green (Invitrogen). We ran samples in technical triplicates and calculated relative mRNA levels using a standard curve and normalized to 36b4 mRNA levels in the same samples.

Histology and immunohistochemistry.

We fixed tissues overnight (16–18 h) in 4% paraformaldehyde in PBS and then dehydrated and embedded them in paraffin. We used 5-μm paraffin sections for standard H&E staining or periodic acid–Schiff staining (Polysciences) following standard procedures or instructions. To quantify average alveoli size, we analyzed pictures from two lungs (9–12 different fields per lung) using MetaMorph software to quantify the pixel numbers in the individual alveoli (defined as empty space bigger than 100 μm2). For immunohistochemistry, we performed antigen retrieval (Vector Lab H-3300) and then used an R.T.U. Vectastain Universal Elite ABC Kit and NovaRED or via-in-pad (VIP) substrate (all from Vector Lab) following the manufacturer′s protocols. The primary antibodies used were goat antibody to Klf2 (Santa Cruz sc-18690), hamster antibody to T1α (Santa Cruz sc-53533), mouse antibody to Cav1 (BD 610406), rabbit antibody to prosurfactant protein B (Millipore AB3432), goat antibody to Sp-c (Santa Cruz M-20) and goat antibody to prosurfactant protein C (Millipore AB3786). The biotin-conjugated secondary antibodies were from Jackson ImmunoResearch.

Light microscopy.

Histology sections were digitally imaged using a fixed, high-resolution Leica DFC420 digital camera mounted on a Leica DMLS microscope equipped with ×4/0.10 C plan, ×10/0.22 C plan, ×20/0.40 C plan, ×40/0.65 N plan and ×100/1.25 oil C plan objective lenses and processed with the Leica Application Suite.

Electron microscopy.

We performed electron microscopy as previously described with minor modifications45. Briefly, samples were fixed in Karnovsky's fixative (4% paraformaldehyde, 2.5% glutaraldehyde and 5 mM CaCl2 in 0.1 M Na cacodylate buffer, pH 7.4) overnight at 4 °C followed by the addition of 1% OsO4 in 0.1 M Na cacodylate buffer, pH 7.4, then en bloc stained with 4% uranyl acetate in 50% ethanol and subsequently dehydrated using a graded series of ethanol solutions followed by propylene oxide and infiltration with epoxy resin (Scipoxy 812, Energy Beam Sciences). After polymerization at 65 °C overnight, we cut thin sections and stained them with uranyl acetate (4% uranyl acetate in 50% ethanol) followed by bismuth subnitrate. We examined sections at an accelerating voltage of 60 kV using a Zeiss EM10C electron microscope.

Isolation of MEFs from SMRTmRID mice.

We derived primary MEFs from E14.5 SMRTmRID heterozygous crossings following standard protocol. Briefly, we decapitated and eviscerated E14.5 embryos, minced the remainder of the carcasses, washed them in DMEM and digested them in 0.25% trypsin plus EDTA at 37 °C for 15 min. We then added DMEM plus 10% FBS media to neutralize trypsin and resuspended the cells after centrifugation and plated them in DMEM plus 10% FBS medium (using a 1 cm × 10 cm dish per embryo). After genotyping, we used MEFs within three passages for transfection experiments.

Liquid chromatography–tandem mass spectrometry quantification of T3 and T4.

We detected and quantified total thyroid hormone levels in serum samples by liquid chromotagrophy–tandem mass spectrometry using calibration curves generated from normal mouse serum using 13C6 T4 (Cambridge Isotope Laboratories) as an internal standard. We generated calibration curves by serial dilution of mouse serum containing known amounts of exogenous T3, rT3 and 13C6 T4 using endogenous T4 as an internal standard and assuming equivalent responses for 12C T4 and 13C6 T4. We added 13C6 T4 to unknown samples of mouse serum (10–20 μl) as an internal standard. Unknown and calibration samples were acidified with formic acid, and protein was precipitated using 2 volumes of CH3CN. We removed precipitated protein by centrifugation (2 min at 8,610g) and diluted the clarified samples with 80 μl HPLC-grade water (JT Baker) before application onto a preconditioned Strata-X SPE column (Phenomenex). Samples were successively washed with 1 ml of 0.2% formic acid, 5% CH3OH (MeOH) in 0.2% formic acid and 40% MeOH in 0.2% formic acid. The thyroid hormones were eluted with 1 ml MeOH, dried under vacuum, reconstituted in 40 μl MeOH and analyzed using an Agilent 1200 HPLC instrument coupled to an LTQ XL linear ion trap mass spectrometer (Thermo Scientific). We resolved samples on a Synergi 30 mm × 2 mm column (Phenomenex) maintained at 10 °C using a H2O-CH3CN–0.1% formic acid gradient (20% CH3CN to 60% CH3CN in 3 min, 60% to 95% CH3CN in 1 min and 95% CH3CN for 3 min) at 0.4 ml min−1. We detected eluted hormones after positive ion electrospray ionization using selective reaction monitoring (SRM) (for T4, the retention time was 4.43 min and SRM was 777.7/731.7; for T3, the retention time was 4.05 min and SRM was 651.7/605.7; for rT3, the retention time was 4.19 min and SRM was 651.7/634.7; and for 13C6 T4, the retention time was 4.43 min and SRM was 783.7/777.7)46,47.

Microarray analysis.

We extracted RNA from E18.5 SMRTmRID and WT littermate control lungs (with or without PTU treatment, each from three embryos), and their purity was assessed by an Agilent Bioanalyzer 2100. We reverse-transcribed 500 ng of RNA into RNA derived from cDNA and biotin–uridine triphosphate labeled it using the Illumina TotalPrep RNA Amplification Kit (Ambion). RNA derived from cDNA was quantified using an Agilent Bioanalyzer 2100 and hybridized to the Illumina MouseRef-8 Expression BeadChip using standard protocols (Illumina). Image data was converted into non-normalized sample probe profiles using the BeadStudio software (Illumina) and analyzed on the VAMPIRE microarray analysis framework48. We constructed stable variance models for each of the two experimental conditions and identified differentially expressed probes using the unpaired VAMPIRE significance test with a two-sided, Bonferroni-corrected threshold (αBonf) of 0.05. The VAMPIRE statistical test is a Bayesian statistical method that computes a model-based estimate of noise at each level of gene expression. This estimate was then used to assess the statistical significance of the apparent differences in gene expression between two experimental conditions. We mapped lists of altered genes generated by VAMPIRE to pathways using the VAMPIRE tool GOby to determine whether any Kyoto Encyclopedia of Genes and Genomes categories were overrepresented using a Bonferroni error threshold of αBonf = 0.05.

Adenoviral infections.

We PCR amplified and cloned mouse Klf2 cDNA into pENTR vector (Invitrogen) and then recombined it into a pAd-CMV vector. We transfected AD293 cells with PacI-digested adenoviral vectors to generate Klf2 adenovirus. GFP adenovirus was a gift from M. Montminy. We used amplified and titered GFP or Klf2 adenovirus to infect (spun at 370g for 10 min at 20–25 °C to increase efficiency) MLE-12 cells at the indicated multiplicity of infection. Under these conditions, approximately 30–50% of cells were transduced, as determined by immunofluorescence microscopy. Forty-eight hours after infection, we harvested the cells for RNA analysis.

Transfections.

We performed transient transfections in CV-1 or MLE-12 cells in triplicate as previously described49 and normalized the luciferase activity to an internal β-galactosidase control. We repeated all transfections at least three times with very similar results, and one representative result is shown in Figure 4 and Supplementary Figures 4 and 10 as the relative luciferase activity ± s.e.m. We amplified the cross-species conserved regions of mouse Pdpn (T1α) promoter (–434 bp to +66 bp), Aqp5 promoter (–635 bp to +10 bp) and Cav1 promoter (–351 bp to +50 bp) by PCR and then cloned them into the NheI-XhoI or NheI-HindIII sites of the pGL4.10 luciferase reporter vector (Promega). We PCR amplified and cloned the conserved Klf2 enhancer region into the BamHI-SalI site of the pGL4.10 luciferase reporter vector. We purchased the pOTB7–hNKX2-1 clone from Open Biosystems. Human NKX2-1 cDNA was PCR amplified from it and cloned into pENTR vector (Invitrogen) and then recombined into pAd-CMV vector.

ChIP.

We followed a previously published protocol for ChIP50. Briefly, we serially fixed 20 million MLE-12 cells with freshly prepared 2 mM disuccinimidyl glutarate (Pierce) for 30 min followed by 1% formaldehyde for 15 min and then quenched them with 125 mM glycine. Nuclei were isolated, lysed in buffer containing 1% sodium dodecyl sulfate, 10 mM EDTA, 50 mM Tris-HCl at pH 8.0 and protease inhibitors (Roche) and sheared with a Diagenode Bioruptor to chromatin fragment sizes of 300–800 bp. Aliquots containing chromatin from 2 million cells were diluted and used for immunoprecipitation overnight at 4 °C with 3 μg rabbit pre-immune IgG (Santa Cruz) or antibodies against Klf2 (Santa Cruz), TR (Thermo Scientific) or SMRT25 (G.B. and R.M.E., unpublished data), with acetylated histone H3 (Millipore) as positive control. Chromatin-antibody complexes were precipitated with protein A/G agarose beads (Millipore), washed, decrosslinked in Chelex 100 (Bio-Rad) and purified using MinElute columns (QIAGEN). Purified DNA was used for qPCR analysis and normalized to input chromatin DNA. The results are shown in Figure 4 and Supplementary Figure 9. as the mean of fold induction of IgG control or percentage of input DNA ± s.e.m. The ChIP primers used are listed in Supplementary Table 4.

Generation of Klf2-chimeric mice.

Klf2−/− ES cells were previously described26. We generated chimeric mice by microinjecting Klf2−/− ES cells into a blastocyst-stage embryo from the C57BL/6J mouse strain and implanting it into the uterus of 2.5 days post coitum pseudopregnant imprinting control region (ICR) recipient females. We recovered and used a total of 38 pups at E18.5 from eight pregnant females for the study.

Statistical analyses.

We used a two-tail unequal variance t test to calculate and determine statistical significance. All data were presented as mean ± s.e.m. unless stated otherwise.

Accession codes.

Microarray data have been deposited in the Gene Expression Omnibus (GEO) with accession code GSE30661.

Accessions

Gene Expression Omnibus

References

  1. 1.

    , , & The branching programme of mouse lung development. Nature 453, 745–750 (2008).

  2. 2.

    Hierarchical organization of lung progenitor cells: is there an adult lung tissue stem cell? Proc. Am. Thorac. Soc. 5, 695–698 (2008).

  3. 3.

    Alveolar type I cells: molecular phenotype and development. Annu. Rev. Physiol. 65, 669–695 (2003).

  4. 4.

    , , , & The great big alveolar TI cell: evolving concepts and paradigms. Cell. Physiol. Biochem. 25, 55–62 (2010).

  5. 5.

    , , , & Knowns and unknowns of the alveolus. Proc. Am. Thorac. Soc. 5, 778–782 (2008).

  6. 6.

    A transcriptional basis for physiology. Nat. Med. 10, 1022–1026 (2004).

  7. 7.

    & A transcriptional co-repressor that interacts with nuclear hormone receptors. Nature 377, 454–457 (1995).

  8. 8.

    et al. Anatomical profiling of nuclear receptor expression reveals a hierarchical transcriptional network. Cell 126, 789–799 (2006).

  9. 9.

    The role of corepressors in transcriptional regulation by nuclear hormone receptors. Annu. Rev. Physiol. 66, 315–360 (2004).

  10. 10.

    & The CoRNR motif controls the recruitment of corepressors by nuclear hormone receptors. Nature 402, 93–96 (1999).

  11. 11.

    & Biological roles and mechanistic actions of co-repressor complexes. J. Cell Sci. 115, 689–698 (2002).

  12. 12.

    , , & Deconstructing repression: evolving models of co-repressor action. Nat. Rev. Genet. 11, 109–123 (2010).

  13. 13.

    & N-CoR-HDAC corepressor complexes: roles in transcriptional regulation by nuclear hormone receptors. Curr. Top. Microbiol. Immunol. 274, 237–268 (2003).

  14. 14.

    et al. SMRT repression of nuclear receptors controls the adipogenic set point and metabolic homeostasis. Proc. Natl. Acad. Sci. USA 105, 20021–20026 (2008).

  15. 15.

    & Preparing for the first breath: genetic and cellular mechanisms in lung development. Dev. Cell 18, 8–23 (2010).

  16. 16.

    Role of transcription factors in fetal lung development and surfactant protein gene expression. Annu. Rev. Physiol. 62, 875–915 (2000).

  17. 17.

    et al. GATA and Nkx factors synergistically regulate tissue-specific gene expression and development in vivo. Development 134, 189–198 (2007).

  18. 18.

    , & Maternal and fetal thyroid function. N. Engl. J. Med. 331, 1072–1078 (1994).

  19. 19.

    et al. The T3Rα gene encoding a thyroid hormone receptor is essential for post-natal development and thyroid hormone production. EMBO J. 16, 4412–4420 (1997).

  20. 20.

    et al. Abnormal heart rate and body temperature in mice lacking thyroid hormone receptor α 1. EMBO J. 17, 455–461 (1998).

  21. 21.

    et al. Ablation of TRα2 and a concomitant overexpression of α1 yields a mixed hypo- and hyperthyroid phenotype in mice. Mol. Endocrinol. 15, 2115–2128 (2001).

  22. 22.

    et al. Genetic analysis reveals different functions for the products of the thyroid hormone receptor α locus. Mol. Cell. Biol. 21, 4748–4760 (2001).

  23. 23.

    et al. Recessive resistance to thyroid hormone in mice lacking thyroid hormone receptor β: evidence for tissue-specific modulation of receptor function. EMBO J. 15, 3006–3015 (1996).

  24. 24.

    & A review of experimental studies of iodine deficiency during fetal development. J. Nutr. 119, 145–151 (1989).

  25. 25.

    et al. Bcl-6 and NF-κB cistromes mediate opposing regulation of the innate immune response. Genes Dev. 24, 2760–2765 (2010).

  26. 26.

    , & Lung Kruppel-like factor, a zinc finger transcription factor, is essential for normal lung development. J. Biol. Chem. 274, 21180–21185 (1999).

  27. 27.

    , & Loss of LKLF function results in embryonic lethality in mice. Transgenic Res. 7, 229–238 (1998).

  28. 28.

    et al. The LKLF transcription factor is required for normal tunica media formation and blood vessel stabilization during murine embryogenesis. Genes Dev. 11, 2996–3006 (1997).

  29. 29.

    et al. The molecular basis of lung morphogenesis. Mech. Dev. 92, 55–81 (2000).

  30. 30.

    , & Genetic disorders influencing lung formation and function at birth. Hum. Mol. Genet. 13(Spec No 2), R207–R215 (2004).

  31. 31.

    & Patterning and plasticity in development of the respiratory lineage. Dev. Dyn. 240, 477–485 (2011).

  32. 32.

    Functions of pulmonary epithelial integrins: from development to disease. Physiol. Rev. 83, 673–686 (2003).

  33. 33.

    , , , & Defects in tracheoesophageal and lung morphogenesis in Nkx2.1−/− mouse embryos. Dev. Biol. 209, 60–71 (1999).

  34. 34.

    et al. SPDEF is required for mouse pulmonary goblet cell differentiation and regulates a network of genes associated with mucus production. J. Clin. Invest. 119, 2914–2924 (2009).

  35. 35.

    et al. Notch signaling controls the balance of ciliated and secretory cell fates in developing airways. Development 136, 2297–2307 (2009).

  36. 36.

    et al. An achaete-scute homologue essential for neuroendocrine differentiation in the lung. Nature 386, 852–855 (1997).

  37. 37.

    et al. Canonical Notch signaling in the developing lung is required for determination of arterial smooth muscle cells and selection of Clara versus ciliated cell fate. J. Cell Sci. 123, 213–224 (2010).

  38. 38.

    et al. Function of the retinoic acid receptors (RARs) during development (II). Multiple abnormalities at various stages of organogenesis in RAR double mutants. Development 120, 2749–2771 (1994).

  39. 39.

    Regulation of fetal lung maturation. Am. J. Physiol. 259, L337–L344 (1990).

  40. 40.

    & Antenatal hormone therapy for fetal lung maturation. Clin. Perinatol. 25, 983–997 (1998).

  41. 41.

    NIH Consensus Development Panel on the Effect of Corticosteroids for Fetal Maturation on Perinatal Outcomes. Effect of corticosteroids for fetal maturation on perinatal outcomes. J. Am. Med. Assoc. 273, 413–418 (1995).

  42. 42.

    & Effects of hormones on fetal lung development. Obstet. Gynecol. Clin. North Am. 31, 949–961 (2004).

  43. 43.

    & Perinatal corticosteroids: a review of research. Part I: antenatal administration. Neonatal Netw. 23, 15–30 (2004).

  44. 44.

    Factors affecting responses of infants with respiratory distress syndrome to exogenous surfactant therapy. Singapore Med. J. 34, 74–77 (1993).

  45. 45.

    et al. Coxsackievirus and adenovirus receptor (CAR) mediates atrioventricular-node function and connexin 45 localization in the murine heart. J. Clin. Invest. 118, 2758–2770 (2008).

  46. 46.

    , & Detection and quantification of 3,5,3′-triiodothyronine and 3,3′,5′-triiodothyronine by electrospray ionization tandem mass spectrometry. J. Am. Soc. Mass Spectrom. 16, 1781–1786 (2005).

  47. 47.

    et al. Free thyroid hormones in serum by direct equilibrium dialysis and online solid-phase extraction–liquid chromatography/tandem mass spectrometry. Clin. Chem. 54, 642–651 (2008).

  48. 48.

    , , & VAMPIRE microarray suite: a web-based platform for the interpretation of gene expression data. Nucleic Acids Res. 33, W627–W632 (2005).

  49. 49.

    et al. NR4A orphan nuclear receptors are transcriptional regulators of hepatic glucose metabolism. Nat. Med. 12, 1048–1055 (2006).

  50. 50.

    , & Protocol for the fast chromatin immunoprecipitation (ChIP) method. Nat. Protoc. 1, 179–185 (2006).

Download references

Acknowledgements

We thank H. Juguilon, M. Karunasiri, S. Kaufman and Y. Dayn for technical support, J. Stubbs, C. Kintner, L. Nagy, S.-H. Hong, J. Jonker and J. Fitzpatrick for helpful discussions, M. Montminy of Salk Institute for GFP adenovirus, J. Simon, L. Grabowski and J. Belcovson for artistic work, and E. Ong and S. Ganley for administrative assistance. We appreciate the help and expertise from M. Wood for the electron microscopy study. We thank J. Codey and the Leona M. and Harry B. Helmsley Charitable Trust for their generous support. L.P. is a Parker B. Francis Fellow supported by the Francis Family Foundation. R.M.E. is an investigator of the Howard Hughes Medical Institute at the Salk Institute for Biological Studies and March of Dimes Chair in Molecular and Developmental Biology. This work was supported by the Howard Hughes Medical Institute and US National Institutes of Health grants 2RO1DK057978, 2RO1HL105278, 5U19DK062434 (R.M.E.) and RO1HL57281 (J.B.L.).

Author information

Affiliations

  1. Howard Hughes Medical Institute, Salk Institute for Biological Studies, La Jolla, California, USA.

    • Liming Pei
    • , Annette Atkins
    •  & Ronald M Evans
  2. Gene Expression Laboratory, Salk Institute for Biological Studies, La Jolla, California, USA.

    • Liming Pei
    • , Mathias Leblanc
    • , Grant Barish
    • , Annette Atkins
    • , Russell Nofsinger
    • , Jamie Whyte
    • , David Gold
    • , Mingxiao He
    • , Kazuko Kawamura
    • , Michael Downes
    • , Ruth T Yu
    •  & Ronald M Evans
  3. Department of Pathology, University of California–San Diego, La Jolla, California, USA.

    • Hai-Ri Li
    •  & Henry C Powell
  4. Department of Molecular Genetics, Biochemistry and Microbiology, College of Medicine, University of Cincinnati, Cincinnati, Ohio, USA.

    • Jerry B Lingrel

Authors

  1. Search for Liming Pei in:

  2. Search for Mathias Leblanc in:

  3. Search for Grant Barish in:

  4. Search for Annette Atkins in:

  5. Search for Russell Nofsinger in:

  6. Search for Jamie Whyte in:

  7. Search for David Gold in:

  8. Search for Mingxiao He in:

  9. Search for Kazuko Kawamura in:

  10. Search for Hai-Ri Li in:

  11. Search for Michael Downes in:

  12. Search for Ruth T Yu in:

  13. Search for Henry C Powell in:

  14. Search for Jerry B Lingrel in:

  15. Search for Ronald M Evans in:

Contributions

L.P. led the project and designed and performed most of the experiments. M.L. is a pathologist who evaluated all the anatomy, histology and staining results. A.A. analyzed the blood T4 and rT3 levels in newborn pups using mass spectrometry. R.N. generated the SMRT knock-in mice. R.T.Y. analyzed the microarray data. H.-R.L. provided expertise in electron microscopy studies. G.B., J.W., D.G., M.H., K.K., M.D., H.C.P. and J.B.L. provided intellectual input and technical expertise. J.B.L. provided the Klf2−/− ES cells. R.M.E. supervised the project. L.P. and R.M.E. wrote the manuscript.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Ronald M Evans.

Supplementary information

PDF files

  1. 1.

    Supplementary Text and Figures

    Supplementary Figures 1–12 and Supplementary Tables 1–4.

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/nm.2450

Further reading