The mechanism by which monosodium urate monohydrate (MSU) crystals intracellularly activate the cryopyrin inflammasome is unknown. The aim of this study was to use a mouse molecular genetics–based approach to test whether the leucine-rich repeat (LRR) domain of cryopyrin is required for MSU crystal–induced inflammation.
Cryopyrin-knockout lacZ (Cryo−Z/−Z) mice and mice with the cryopyrin LRR domain deleted and fused to the lacZ reporter (CryoΔLRR Z/ΔLRR Z) were generated using bacterial artificial chromosome–based targeting vectors, which allow for large genomic deletions. Bone marrow–derived macrophages from CryoΔLRR Z/ΔLRR Z mice, Cryo−Z/−Z mice, and congenic wild-type (WT) mice were challenged with endotoxin-free MSU crystals under serum-free conditions. Phagocytosis and cytokine expression were assessed by flow cytometry and enzyme-linked immunosorbent assay. MSU crystals also were injected into mouse synovial-like subcutaneous air pouches. The in vivo inflammatory responses were examined.
Release of interleukin-1β (IL-1β), but not CXCL1 and tumor necrosis factor α, was impaired in CryoΔLRR Z/ΔLRR Z and Cryo−Z/−Z mouse bone marrow–derived macrophages compared with WT mouse bone marrow–derived macrophages in response to not only MSU crystals but also other known stimuli that activate the cryopyrin inflammasome. In addition, a comparable percentage of MSU crystals taken up by each type of bone marrow–derived macrophage was observed. Moreover, total leukocyte infiltration in the air pouch and IL-1β production were attenuated in Cryo−Z/−Z and CryoΔLRR Z/ΔLRR Z mice at 6 hours postinjection of MSU crystals compared with WT mice.
MSU crystal–induced inflammatory responses were comparably attenuated both in vitro and in vivo in CryoΔLRR Z/ΔLRR Z and Cryo−Z/−Z mice. Hence, the LRR domain of cryopyrin plays a role in mediating MSU crystal–induced inflammation in this model.
In gout, the deposition of monosodium urate monohydrate (MSU) crystals in articular joints and bursal tissues can be asymptomatic or can be associated with the pathogenesis of acute, episodic, self-limiting joint inflammation (1–3). The interaction of MSU crystals with resident cells such as synovial lining cells and macrophages in the joint is believed to be the primary trigger for the intense neutrophil ingress that drives episodes of gouty arthritis (4). Cells encountering MSU crystals express a broad array of inflammation mediators that drive and amplify acute gouty inflammation, including arachidonate metabolites, the cytokines interleukin-1β (IL-1β), tumor necrosis factor α (TNFα), CXCL1 (GROα), and CXCL8 (IL-8) (5–9) and the calgranulins S100A8 and S100A9 (10).
The naked MSU crystal has a negatively charged, highly reactive surface that nonspecifically binds at least 25 different serum proteins (11) and also binds plasma membrane proteins including certain integrins (12, 13). MSU crystal binding of C5 and C5 catalysis on the crystal surface (14) promote C5b–C9 membrane attack complex assembly that contributes to both intraarticular CXCL8 expression and acute neutrophilic inflammation in experimental MSU crystal–induced knee arthritis (15). Several studies have demonstrated the importance of innate immunity in acute gouty inflammation. MSU crystals can functionally engage the canonical signaling pathway from Toll-like receptor 2 (TLR-2) to NF-κB activation mediated by the shared TLR and IL-1 receptor adaptor protein myeloid differentiation factor 88 (MyD88) (16). TLR-2 and TLR-4 each mediate macrophage uptake of the MSU crystal in vitro and MSU crystal–induced inflammation in vivo (17). In addition, MyD88 plays a major role in macrophage uptake of the MSU crystal and is essential for MSU crystal–induced inflammation in vivo (17). Moreover, direct engagement of CD14, which is a shared TLR-2 and TLR-4 adaptor molecule, is a major determinant of the inflammatory potential of the MSU crystals (18). Furthermore, the cytoplasmic cryopyrin (also known as NALP3 or NLRP3) inflammasome complex, which is principally expressed in phagocytes, is pivotal for acute MSU crystal–induced inflammation (19).
MSU crystals appear to be among the stimuli that trigger aggregation and activation of the cryopyrin inflammasome through pyrin–pyrin domain interactions of cryopyrin and of the adaptor protein apoptosis-associated speck-like protein containing a caspase activation and recruitment domain (ASC). Activation of the inflammasome complex results in the recruitment and proteolytic cleavage of caspase 1 (19). In this context, MSU crystal–induced caspase 1 activation, and subsequent cleavage, maturation, and release of IL-1β, are markedly decreased in macrophages from mice deficient in cryopyrin, ASC, or caspase 1 in vitro (19). Moreover, MSU crystal–induced peritoneal neutrophil influx is blunted in mice deficient in cryopyrin, ASC, or caspase 1 (19, 20).
Similar to the structure of the TLR domain, cryopyrin has an LRR domain at its C-terminus that is also proposed to be a ligand-sensing motif (21). In this model, cryopyrin is normally present in the cytoplasm in an inactive form but becomes active when the LRR domain is engaged by an agonist. This is thought to be attributable to the conformational rearrangement of this molecule, which exposes the oligomerization domain (nucleotide-binding site/NACHT) and subsequently the effector domain (pyrin-binding domain) (21). In this study, we investigated whether the LRR domain of cryopyrin is required for MSU crystal–induced inflammation, using a novel recombinant mouse with the cryopyrin LRR domain deleted and fused to the lacZ reporter (CryoΔLRR Z/ΔLRR Z). Macrophages from these mice stimulated in vitro do not induce caspase 1 activation and IL-1β release. The in vivo inflammatory response in subcutaneous air pouches in these mice is significantly reduced. In addition, the IL-1β release in CryoΔLRR Z/ΔLRR Z mouse macrophages in vitro in response to several other known cryopyrin activators is also decreased markedly.
MATERIALS AND METHODS
All chemical reagents were obtained from Sigma-Aldrich, unless otherwise indicated. Triclinic MSU crystals were prepared under pyrogen-free conditions, using uric acid pretreated for 2 hours at 200°C prior to crystallization (17). The crystals were suspended at 25 mg/ml in sterile, endotoxin-free phosphate buffered saline and verified to be free of detectable lipopolysaccharide contamination (<0.025 endotoxin units/ml) by the Limulus amebocyte cell lysate test (BioWhittaker). Peptidoglycan and R837 were obtained from Invitrogen, and bacterial RNA was obtained from Ambion.
Mice were generated at Regeneron Pharmaceuticals using the VelociGene approach, as previously described (22). This approach has been useful for studying in situ expression of targeted proteins, particularly when specific antisera are unavailable. A targeting vector was constructed that included an in-frame reporter lacZ gene cloned next to an out-of-frame neomycin resistance gene flanked by loxP sites and driven by a promoter that allowed for positive selection in both bacterial and mammalian cells. The LacZ-Neo cassette was ligated to double-stranded oligonucleotides and used for the generation of bacterial artificial chromosome–based targeting vectors lacking the cryopyrin LRR domain (CryoΔLRR Z/ΔLRR Z) or deficient in cryopyrin (Cryo−Z/−Z), as shown in Figure 1B. These constructs were microinjected into embryonic stem cells derived from the (129/Sv × C57BL/6)F1 mouse background to allow for proper recombination. Correctly targeted embryonic stem cells carrying the targeting construct were injected into BALB/c mouse blastocysts, which were then implanted into pseudopregnant CD1 mouse foster mothers. Male chimeras were bred with C57BL/6 mice to screen for germline-transmitted offspring. Mice bearing the targeted allele were screened by polymerase chain reaction (PCR).
To confirm gene expression of the “truncated” cryopyrin (ΔLRR), RNA isolated from the bone marrow of these mice (using TRIzol) was subjected to reverse transcription–PCR (ABI TaqMan) using the following exonic primer pairs: 5′-CGAGAAAGGCTGTATCCCAG-3′ and 5′-GCTAGGATGGTTTTCCCGAT-3′ (exons 1–3), 5′-CACGTGGTTTCCTCCTTTTG-3′ and 5′-TGGTGAAGGAGGGCTTGATA-3′ (exons 3–9), 5′-CACGTGGTTTCCTCCTTTTG-3′ and 5′-TTGACTGTAGCGGCTGATGTTG-3′ (exon 3 to lacZ), 5′-GGTAAACTGGCTCGGATTAGGG-3′ and 5′-TTGACTGTAGCGGCTGATGTTG-3′ (lacZ to lacZ), and 5′-GGTCTTACTCCTTGGAGGCCATGT-3′ and 5′-GACCCCTTCATTGACCTCAACTACA-3′ (GAPDH). Protein expression of the truncated cryopyrin (ΔLRR) was also confirmed by Western blot analysis on bone marrow–derived macrophages from these mice, using antibodies to either β-galactosidase (β-gal; Invitrogen) or cryopyrin N14 (Santa Cruz Biotechnology).
Isolation and culture of murine macrophages.
All animal experiments were conducted in a humane manner according to institutionally approved protocols. Background-matched wild-type (WT), Cryo−Z/−Z, and CryoΔLRR Z/ΔLRR Z mice were backcrossed at least 4 generations on a C57BL/6 background and were maintained under specific pathogen–free conditions and genotyped by PCR. Bone marrow–derived macrophages were prepared from 8–10-week-old homozygous Cryo−Z/−Z and CryoΔLRR Z/ΔLRR Z mice as well as congenic WT control mice.
Western blot analysis and immunoprecipitation.
Bone marrow–derived macrophages were lysed with buffer containing 50 mM Tris pH 7.8, 50 mM NaCl, 0.1% Nonidet P40, 5 mM EDTA, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitors (Roche) on ice and passed through a 22-gauge needle 10 times. For the immunoprecipitation experiment, cell lysates were incubated with anticryopyrin (N-14) peptide antibody (Santa Cruz Biotechnology) at 4°C overnight, and Protein G–Sepharose Fast Flow (Sigma) was added for 2 hours at 4°C. Beads were spun down and washed 3 times with the same buffer. The washed beads or cell lysates were separated on 4–15% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gels and then transferred to polyvinylidene difluoride membranes. The membranes were then used for immunoblot analyses with the antibodies indicated. The same amount of conditioned media was subjected to SDS-PAGE and Western blot analyses of caspase 1 and IL-1β expression, using antibodies to caspase 1 (BioVision) and IL-1β (BioVision).
Assays of phagocytosis and cytokine production.
Bone marrow–derived macrophages of individual genotypes were treated with MSU crystals (0.5 mg/ml) for 2 hours at 37°C and then were washed 3 times with cold PBS containing 5 mM EDTA and harvested in the same buffer. The proportion of the macrophages taking up MSU crystals was assessed by flow cytometric analysis based on increased side scatter (23).
We evaluated generation of IL-1β and CXCL1 by DuoSet enzyme-linked immunosorbent assay (ELISA; R&D Systems), following the manufacturer's protocol, by testing conditioned media collected from mouse bone marrow–derived macrophages (5 × 105/well) stimulated with MSU crystals (0.5 mg/ml) for 24 hours.
Studies of synovial-like subcutaneous air pouches.
Subcutaneous pouches were generated by the injection of sterile-filtered air to generate an accessible space that developed a synovium-like membrane within 7 days, as previously described (17). Briefly, anesthetized 8–10-week-old WT, Cryo−Z/−Z, and CryoΔLRR Z/ΔLRR Z mice were injected with 5 ml of sterile air into the subcutaneous tissue of the back, followed by a second injection of 3 ml of sterile air into the pouch 3 days later. MSU crystals (3 mg) in 1 ml of sterile endotoxin-free PBS were injected into the pouch 7 days after the first injection of air. The mice were killed, and pouch fluids were harvested at specific time points by injecting 5 ml of PBS containing 5 mM EDTA, and cells infiltrating into the air pouch were counted manually using a hemocytometer. Smears of cells from the air pouches on the slides were prepared by centrifugation of 105 cells in cytofunnels (Thermo Shandon) in a Cytospin 4 centrifuge (Thermo Shandon) at 110g for 2 minutes. Leukocyte population counts were measured via Wright-Giemsa staining of cytospin slides. IL-1β expression was determined by ELISA, as described above, in supernatants of air pouch exudates.
Data are presented as the mean ± SD. Statistical analyses were performed using Student's 2-tailed t-test.
Generation and characterization of CryoΔLRR Z/ΔLRR Z mice.
To investigate whether the cryopyrin LRR domain plays a role in the inflammatory response, we generated cryopyrin LRR deletion mutant mice (CryoΔLRR Z/ΔLRR Z), as well as cryopyrin-knockout (Cryo−Z/−Z) mice (Figures 1A–C) as described in Materials and Methods. To examine expression of the recombinant gene, we first isolated RNA from the bone marrow of these mice and performed reverse transcription, followed by PCR using primers derived from exonic sequence coding for various domains. We confirmed that the “truncated” cryopyrin LRR mutant was expressed at the RNA level (Figure 1D). Next, we isolated cell lysates from bone marrow–derived macrophages from WT, CryoΔLRR Z/ΔLRR Z, and Cryo−Z/−Z mice and measured protein expression by Western blotting using an antibody to β-gal. Expression of the truncated cryopyrin (ΔLRR) fused to β-gal and expression of β-gal alone in cryopyrin-knockout mice in which the entire gene was replaced by β-gal were observed (Figure 1E). Immunoprecipitation analyses with a cryopyrin-specific antibody also confirmed expression of the truncated cryopyrin (ΔLRR) fused to β-gal in CryoΔLRR Z/ΔLRR Z mice (data not shown), as well as expression of cryopyrin in WT mice (data not shown). These mice were viable and fertile, the pups were born at the expected Mendelian ratio, and there was no apparent phenotype.
Attenuation of MSU crystal–induced inflammatory responses in Cryo−Z/−Z mice.
First, bone marrow–derived macrophages were generated from Cryo−Z/−Z mice and congenic WT mice. As with our previous studies (17, 18), to avoid potential masking effects of both serum protein opsonization of the crystals (1) and of crystal-induced complement activation (14, 15), bone marrow–derived macrophages were treated with endotoxin-free MSU crystals at a concentration of 0.5 mg/ml under entirely serum-free conditions. At 24 hours, MSU crystals induced release of IL-1β and CXCL1 (Figures 2A and B, respectively) in WT mouse bone marrow–derived macrophages. However, release of IL-1β (Figure 2A), but not of CXCL1 (Figure 2B) and TNFα (data not shown) was blunted in Cryo−Z/−Z bone marrow–derived macrophages in response to MSU crystals. Notably, there was no impairment in the uptake of MSU crystals in Cryo−Z/−Z mouse bone marrow–derived macrophages after 2 hours stimulation at 37°C (Figure 2C), suggesting that cryopyrin is not involved in phagocytosis of MSU crystals. In vivo studies using the air pouch model revealed that MSU crystal–induced leukocyte infiltration peaked at 6 hours postinjection of MSU crystals and was self-limiting by 24 hours postinjection of MSU crystals in the air pouch of WT mice (Figure 2D). In contrast, the total number of leukocytes infiltrated in the air pouch of Cryo−Z/−Z mice was markedly suppressed at 6 hours postinjection (Figure 2D). This result was inconsistent with previous observations in the peritonitis model (20).
Impaired MSU crystal–induced inflammatory responses in CryoΔLRR Z/ΔLRR Z mice.
In vitro and in vivo studies similar to those described in Figure 2 were carried out in CryoΔLRR Z/ΔLRR Z mice. As shown in Figures 3A and B, MSU crystal–induced release of IL-1β, but not CXCL1, was blunted in bone marrow–derived macrophages of CryoΔLRR Z/ΔLRR Z mice in vitro, comparable with that in bone marrow–derived macrophages from Cryo−Z/−Z mice. In vivo air pouch model experiments demonstrated that the MSU crystal–induced leukocyte infiltration observed in WT mice at 6 hours postinjection was markedly decreased in CryoΔLRR Z/ΔLRR Z mice (Figure 3C). In addition, there was a significant decrease in IL-1β release in the air pouch of CryoΔLRR Z/ΔLRR Z mice compared with that in the air pouch of WT mice (Figure 3D). These data suggest that the cryopyrin LRR domain is required for mediating MSU crystal–induced inflammatory responses.
Attenuation of caspase 1 activation and IL-1β cleavage in CryoΔLRR Z/ΔLRR Z mouse bone marrow–derived macrophages in response to MSU crystals in vitro.
Next, we examined caspase 1 activation in bone marrow–derived macrophages from CryoΔLRR Z/ΔLRR Z mice in response to MSU crystals in vitro. As depicted in the top panel of Figure 4, the MSU crystal–induced caspase 1 activation observed in WT mouse bone marrow–derived macrophages was diminished in bone marrow–derived macrophages of not only Cryo−Z/−Z mice but also CryoΔLRR Z/ΔLRR Z mice. Similarly, IL-1β release was repressed in bone marrow–derived macrophages from Cryo−Z/−Z mice and CryoΔLRR Z/ΔLRR Z mice in response to MSU crystals, compared with that in WT mouse bone marrow–derived macrophages (bottom panel of Figure 4). This suggests that the LRR domain of cryopyrin is needed for MSU crystal–induced caspase 1 activation and IL-1β release in bone marrow–derived macrophages.
Impaired IL-1β release in CryoΔLRR Z/ΔLRR Z mouse bone marrow–derived macrophages in response to several other known cryopyrin activators in vitro.
To determine whether the cryopyrin LRR domain is generally required for mediating IL-1β release in response to inflammatory stimuli, we examined IL-1β release in bone marrow–derived macrophages from CryoΔLRR Z/ΔLRR Z mice and compared it with that in WT and Cryo−Z/−Z mouse bone marrow–derived macrophages in response to several stimuli known to activate cryopyrin, such as peptidoglycan, bacterial RNA, R837, and crude LPS. As seen in Figure 5A, IL-1β release was induced by all of these stimuli in WT mouse bone marrow–derived macrophages but was reduced significantly in bone marrow–derived macrophages from CryoΔLRR Z/ΔLRR Z mice, comparable with that in bone marrow–derived macrophages from Cryo−Z/−Z mice. In contrast, there was no significant difference in CXCL1 release in all types of bone marrow–derived macrophages in response to all of the stimuli (Figure 5B). These data suggest that the cryopyrin LRR domain is essential for mediating IL-1β release in response to inflammatory stimuli known to activate the cryopyrin inflammasome.
In this study, we demonstrated that the novel CryoΔLRR Z/ΔLRR Z mice have decreased MSU crystal–induced caspase 1 activation and IL-1β release in bone marrow–derived macrophages in vitro and leukocyte infiltration in the air pouch model in vivo. In addition, we showed that CryoΔLRR Z/ΔLRR Z mice have decreased IL-1β release in bone marrow–derived macrophages in response to several other known cryopyrin activators in vitro.
The cryopyrin inflammasome is activated by several pathogen-associated molecular patterns (PAMPs), including bacterial muramyl dipeptide (MDP), a degradation product of the bacterial cell wall component peptidoglycan, the microbial toxins, RNA of bacterial and viral origin, and cytosolic microbial and host DNA (24–27), as well as danger-associated molecular patterns (DAMPs) such as ATP, imidazoquinoline, MSU crystals, asbestos, and silica (25, 28–33). It is still not clear how cryopyrin senses these diverse activators to trigger inflammasome complex formation that leads to caspase 1 activation and IL-1β release. One putative mechanism is that each of the activators directly or indirectly interacts with the LRR domain of cryopyrin, leading to a conformational change in cryopyrin and subsequently to inflammasome assembly. MDP was recently demonstrated to directly bind to recombinant NALP1, and MDP interaction with the LRR region of NALP1 is essential for caspase 1 activation mediated by the reconstituted NALP1 inflammasome (34). These data indicate that NALPs could directly interact with their activators. Although we do not yet know whether MSU crystal activation occurs via direct interaction with the LRR domain of cryopyrin, the results of our study suggest a role for the LRR domain of cryopyrin in MSU crystal–induced inflammatory responses.
Interestingly, recent findings showed that potassium efflux, lowering intracellular potassium levels, is a common requirement for cryopyrin inflammasome activation triggered by all known activators including MSU crystals (33, 35, 36). In addition to potassium efflux, reactive oxygen species production is a necessary step in activation of the cryopyrin inflammasome (32, 33). It is unlikely that each of the activators is “specifically” recognized by cryopyrin. Dostert et al proposed that these activators could induce a cellular stress situation that in all cases results in modification of ≥1 membrane-associated proteins, which then trigger a signaling cascade leading to activation of cryopyrin (32). Because we observed that IL-1β release in bone marrow–derived macrophages of CryoΔLRR Z/ΔLRR Z mice was impaired in response to several known activators of cryopyrin in vitro, it is possible that these activators may activate protein(s) that directly interact(s) with the cryopyrin LRR domain.
The chaperone heat-shock protein 90 (Hsp90) and the co–chaperone-like, ubiquitin ligase–associated protein SGT1 have been shown to bind the LRR domain of cryopyrin, which is essential for the function of cryopyrin inflammasome (37). These proteins maintain cryopyrin in an inactive but signaling-competent state and disassociate from cryopyrin once activating signals are detected, thereby allowing conformational change of cryopyrin that enables the interaction of cryopyrin with other components such as ASC and procaspase 1. In the absence of Hsp90, cryopyrin becomes unstable and is degraded by the proteosome (37). Thus, if cryopyrin is missing the LRR domain, the cryopyrin inflammasome will lose function, because Hsp90 and SGT1 cannot interact with cryopyrin. In addition, cryopyrin without the LRR domain may be unstable and therefore unable to form an inflammasome complex. This may explain why IL-1β release was impaired in bone marrow–derived macrophages from CryoΔLRR Z/ΔLRR Z mice in response to MSU crystals and several other known activators of cryopyrin in vitro. At present, we are not sure whether this result is attributable to instability of the truncated cryopyrin.
The LRR domain of human cryopyrin has been shown to be alternatively spliced in the 3′ region of the gene, resulting in a large number of cryopyrin forms of differing lengths and LRR composition (38). One of the most common alternative splice forms expressed in human leukocytes lacks most of the LRR domain, which is similar to the CryoΔLRR Z/ΔLRR Z studied here in mice. It is still unclear whether these alternatively spliced forms are expressed at the protein level or whether these forms have any unique function.
It is also unclear whether cryopyrin alternative splicing occurs in mice. According to one proposed model, the LRR domain serves an inhibitory role in its native state by preventing inflammasome oligomerization. In vitro evidence suggests that expression of a truncated cryopyrin results in a constitutively active cryopyrin inflammasome (39, 40), which could have a similar effect as the gain-of-function mutations observed in patients with cryopyrinopathies. However, interpretation of these in vitro models may be limited. This is due to the fact that 3 different components of the inflammasome (cryopyrin, ASC, caspase 1) and proIL-1β have to be simultaneously introduced into a cell line (which is normally a non–myeloid cell line such as 293 cells) via transfection.
The efficiency of expression of all 4 components may be inconsistent from cell to cell, with some cells expressing only some of the components due to the difficulty of simultaneous transfection of 4 different complementary DNA constructs. In addition, the expression level of each component is artificial and may not truly reflect the endogenous level of each component in myeloid cells. These limitations prompted us to study the absence of the LRR domain in an in vivo mouse model.
In contrast, CryoΔLRR Z/ΔLRR Z mice are phenotypically normal, similar to the Cryo−Z/−Z mice, and our in vitro mouse studies do not support the previous results observed in transfected human cell lines. Our data are more consistent with the LRR domain having a functional role in PAMP or DAMP sensing, a structural role in cryopyrin protein stability, or a contributory role in inflammasome protein–protein interactions. The primary limitation of this mouse model is the construct design, which included a LacZ fusion protein in place of the LRR domain. This approach was chosen in order to confirm expression of the protein and study tissue distribution of the ΔLRR form of cryopyrin. However, LacZ is a relatively large protein, and evidence suggests that oligomerization of multiple cryopyrin monomers, adaptor proteins, and chaperone proteins is crucial to inflammasome function (24, 37, 40). Therefore, it is possible that the LacZ fusion product in the CryoΔLRR Z/ΔLRR Z mice interferes with inflammasome oligomerization, resulting in the observed null phenotype.
The role of the LRR domain of cryopyrin in MSU crystal–induced inflammatory responses suggested by our studies in CryoΔLRR Z/ΔLRR Z mice makes it a potential drug target for gouty inflammation. Because the LRR domain of cryopyrin has a potential to directly or indirectly engage with a vast array of structurally unrelated PAMPs or DAMPs to activate the cryopyrin inflammasome leading to innate immune inflammatory responses, targeting the LRR domain of cryopyrin has a huge drug potential for host defense.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Liu-Bryan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Hoffman, Liu-Bryan.
Acquisition of data. Scott, Mueller, Misaghi, Stevens, Yancopoulos, Murphy, Valenzuela.
Analysis and interpretation of data. Hoffman, Scott, Mueller, Misaghi, Stevens, Yancopoulos, Murphy, Valenzuela, Liu-Bryan.